advanced training.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Check for Colab Env\n",
    "> if we are in colab env we need to get the data + scripts from github or somewhere"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting opencv-python\n",
      "  Downloading opencv_python-4.6.0.66-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (60.9 MB)\n",
      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m0.4/60.9 MB\u001b[0m \u001b[31m57.4 kB/s\u001b[0m eta \u001b[36m0:17:34\u001b[0mm"
     ]
    }
   ],
   "source": [
    "import os\n",
    "if \"COLAB_GPU\" in os.environ:\n",
    "    _useMultiProcessing = False\n",
    "    os.system(\"git clone https://github.com/zaxo7/UNet-Playground.git\")\n",
    "    os.chdir(\"/content/UNet-Playground/\")\n",
    "else:\n",
    "    _useMultiProcessing = True\n",
    "    \n",
    "if \"PAPERSPACE_FQDN\" in os.environ:\n",
    "    #!apt update\n",
    "    #!apt install libgl1 -y\n",
    "    !pip install opencv-python \n",
    "    #!git pull\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- enable module autoreload and fix an import bug"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-06-05T08:49:33.834263Z",
     "iopub.status.busy": "2022-06-05T08:49:33.834047Z",
     "iopub.status.idle": "2022-06-05T08:49:33.854206Z",
     "shell.execute_reply": "2022-06-05T08:49:33.853540Z",
     "shell.execute_reply.started": "2022-06-05T08:49:33.834263Z"
    }
   },
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "import sys,os\n",
    "sys.path.append(os.getcwd() + '/scripts')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- import the scripts and data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-06-05T08:49:38.584387Z",
     "iopub.status.busy": "2022-06-05T08:49:38.583620Z",
     "iopub.status.idle": "2022-06-05T08:49:38.626554Z",
     "shell.execute_reply": "2022-06-05T08:49:38.625685Z",
     "shell.execute_reply.started": "2022-06-05T08:49:38.583814Z"
    }
   },
   "outputs": [],
   "source": [
    "from scripts import data, model\n",
    "import glob\n",
    "import matplotlib.pyplot as plt\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "import cv2\n",
    "import pandas as pd\n",
    "import time\n",
    "from scripts.DataGenerator import DataGenerator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-05-23T13:34:08.419387Z",
     "iopub.status.busy": "2022-05-23T13:34:08.419110Z",
     "iopub.status.idle": "2022-05-23T13:34:08.691432Z",
     "shell.execute_reply": "2022-05-23T13:34:08.690526Z",
     "shell.execute_reply.started": "2022-05-23T13:34:08.419359Z"
    }
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- model parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-06-04T23:45:15.774859Z",
     "iopub.status.busy": "2022-06-04T23:45:15.774082Z",
     "iopub.status.idle": "2022-06-04T23:45:15.817691Z",
     "shell.execute_reply": "2022-06-04T23:45:15.816844Z",
     "shell.execute_reply.started": "2022-06-04T23:45:15.774830Z"
    }
   },
   "outputs": [],
   "source": [
    "_batch_size = 8\n",
    "_epochs = 200\n",
    "_workers = 8\n",
    "_model_name = f'RBC_BinaryCrossentropy_data2_data2_60'\n",
    "images_per_epoch = 1000\n",
    "train_steps = images_per_epoch//_batch_size\n",
    "#validation_steps = len(test_img_tile_files)//_batch_size\n",
    "\n",
    "Train = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-06-04T23:45:15.819986Z",
     "iopub.status.busy": "2022-06-04T23:45:15.818744Z",
     "iopub.status.idle": "2022-06-04T23:45:16.812716Z",
     "shell.execute_reply": "2022-06-04T23:45:16.811612Z",
     "shell.execute_reply.started": "2022-06-04T23:45:15.819943Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "skipping empty images with keep probability of 0.05\n",
      "before skip we have 6640 images\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 6640/6640 [00:00<00:00, 35112.27it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "after skip we have 4832 images\n",
      "skipping empty images with keep probability of 0.05\n",
      "before skip we have 1834 images\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1834/1834 [00:00<00:00, 15192.40it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "after skip we have 1312 images\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "database = \"data\"\n",
    "\n",
    "\n",
    "train_img_tile_files = sorted(glob.glob(f'{database}/train/tiles/*.jpg'))\n",
    "train_mask_tile_label_files = sorted(glob.glob(f'{database}/train/mask/tiles/*.jpg'))\n",
    "train_edge_tile_label_files = sorted(glob.glob(f'{database}/train/edge/tiles/*.jpg'))\n",
    "\n",
    "database = \"data\"\n",
    "\n",
    "test_img_tile_files = sorted(glob.glob(f'{database}/test/tiles/*.jpg'))\n",
    "test_mask_tile_label_files = sorted(glob.glob(f'{database}/test/mask/tiles/*.jpg'))\n",
    "test_edge_tile_label_files = sorted(glob.glob(f'{database}/test/edge/tiles/*.jpg'))\n",
    "\n",
    "train_gen = DataGenerator(train_img_tile_files, train_mask_tile_label_files, train_edge_tile_label_files, _batch_size, skip_empty = True, noise_prob = 0)\n",
    "\n",
    "test_gen = DataGenerator(test_img_tile_files, test_mask_tile_label_files, test_edge_tile_label_files, _batch_size, skip_empty = True, augment=False)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- load weights and compile it"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-06-04T23:45:16.814574Z",
     "iopub.status.busy": "2022-06-04T23:45:16.814304Z",
     "iopub.status.idle": "2022-06-04T23:45:21.659916Z",
     "shell.execute_reply": "2022-06-04T23:45:21.659118Z",
     "shell.execute_reply.started": "2022-06-04T23:45:16.814548Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "loading model RBC_BinaryCrossentropy_data2_data2_60\n",
      "66\n",
      "                Layer Name          Output Shape  Layer Trainable\n",
      "0                  input_4   (None, 188, 188, 3)             True\n",
      "1                conv2d_45  (None, 186, 186, 96)             True\n",
      "2   batch_normalization_54  (None, 186, 186, 96)             True\n",
      "3            activation_54  (None, 186, 186, 96)             True\n",
      "4                conv2d_46  (None, 186, 186, 32)             True\n",
      "5   batch_normalization_55  (None, 186, 186, 32)             True\n",
      "6            activation_55  (None, 186, 186, 32)             True\n",
      "7                conv2d_47  (None, 184, 184, 32)             True\n",
      "8   batch_normalization_56  (None, 184, 184, 32)             True\n",
      "9            activation_56  (None, 184, 184, 32)             True\n",
      "10         max_pooling2d_9    (None, 92, 92, 32)             True\n",
      "11               conv2d_48    (None, 90, 90, 64)             True\n",
      "12  batch_normalization_57    (None, 90, 90, 64)             True\n",
      "13           activation_57    (None, 90, 90, 64)             True\n",
      "14               conv2d_49    (None, 88, 88, 64)             True\n",
      "15  batch_normalization_58    (None, 88, 88, 64)             True\n",
      "16           activation_58    (None, 88, 88, 64)             True\n",
      "17        max_pooling2d_10    (None, 44, 44, 64)             True\n",
      "18               conv2d_50   (None, 42, 42, 128)             True\n",
      "19  batch_normalization_59   (None, 42, 42, 128)             True\n",
      "20           activation_59   (None, 42, 42, 128)             True\n",
      "21               conv2d_51   (None, 40, 40, 128)             True\n",
      "22  batch_normalization_60   (None, 40, 40, 128)             True\n",
      "23           activation_60   (None, 40, 40, 128)             True\n",
      "24        max_pooling2d_11   (None, 20, 20, 128)             True\n",
      "25               conv2d_52   (None, 18, 18, 256)             True\n",
      "26  batch_normalization_61   (None, 18, 18, 256)             True\n",
      "27           activation_61   (None, 18, 18, 256)             True\n",
      "28               conv2d_53   (None, 16, 16, 256)             True\n",
      "29  batch_normalization_62   (None, 16, 16, 256)             True\n",
      "30           activation_62   (None, 16, 16, 256)             True\n",
      "31      conv2d_transpose_9   (None, 32, 32, 128)             True\n",
      "32  batch_normalization_63   (None, 32, 32, 128)             True\n",
      "33            cropping2d_9   (None, 32, 32, 128)             True\n",
      "34           activation_63   (None, 32, 32, 128)             True\n",
      "35           concatenate_9   (None, 32, 32, 256)             True\n",
      "36               conv2d_54   (None, 30, 30, 128)             True\n",
      "37  batch_normalization_64   (None, 30, 30, 128)             True\n",
      "38           activation_64   (None, 30, 30, 128)             True\n",
      "39               conv2d_55   (None, 28, 28, 128)             True\n",
      "40  batch_normalization_65   (None, 28, 28, 128)             True\n",
      "41           activation_65   (None, 28, 28, 128)             True\n",
      "42     conv2d_transpose_10    (None, 56, 56, 64)             True\n",
      "43  batch_normalization_66    (None, 56, 56, 64)             True\n",
      "44           cropping2d_10    (None, 56, 56, 64)             True\n",
      "45           activation_66    (None, 56, 56, 64)             True\n",
      "46          concatenate_10   (None, 56, 56, 128)             True\n",
      "47               conv2d_56    (None, 54, 54, 64)             True\n",
      "48  batch_normalization_67    (None, 54, 54, 64)             True\n",
      "49           activation_67    (None, 54, 54, 64)             True\n",
      "50               conv2d_57    (None, 52, 52, 64)             True\n",
      "51  batch_normalization_68    (None, 52, 52, 64)             True\n",
      "52           activation_68    (None, 52, 52, 64)             True\n",
      "53     conv2d_transpose_11  (None, 104, 104, 32)             True\n",
      "54  batch_normalization_69  (None, 104, 104, 32)             True\n",
      "55           cropping2d_11  (None, 104, 104, 32)             True\n",
      "56           activation_69  (None, 104, 104, 32)             True\n",
      "57          concatenate_11  (None, 104, 104, 64)             True\n",
      "58               conv2d_58  (None, 102, 102, 32)             True\n",
      "59  batch_normalization_70  (None, 102, 102, 32)             True\n",
      "60           activation_70  (None, 102, 102, 32)             True\n",
      "61               conv2d_59  (None, 100, 100, 32)             True\n",
      "62  batch_normalization_71  (None, 100, 100, 32)             True\n",
      "63           activation_71  (None, 100, 100, 32)             True\n",
      "64                    mask   (None, 100, 100, 1)             True\n",
      "65                    edge   (None, 100, 100, 1)             True\n"
     ]
    }
   ],
   "source": [
    "print(f\"loading model {_model_name}\")\n",
    "#unet.load_weights(f\"models/{_model_name}_best.h5\")\n",
    "\n",
    "#unet.save('./tmp_model',save_format='tf')\n",
    "\n",
    "objects = {\n",
    "    \"tversky_loss\":model.tversky_loss,\n",
    "    \"mean_iou\": model.mean_iou,\n",
    "    \"dsc\":model.dsc,\n",
    "    \"tversky\":model.tversky,\n",
    "    \"tversky_b\":model.tversky_b,\n",
    "    \"acc\": tf.keras.metrics.Accuracy,\n",
    "    \"tversky_loss_b\": model.tversky_loss_b\n",
    "}\n",
    "\n",
    "\n",
    "\n",
    "unet = tf.keras.models.load_model(f'./models/{_model_name}', custom_objects=objects)\n",
    "\n",
    "print(len(unet.layers))\n",
    "\n",
    "layers = []\n",
    "#freeze the layers\n",
    "#freeze the all the layers\n",
    "#layers = range(0, 66)\n",
    "#freeze the edge layer\n",
    "#layers = [-1]\n",
    "for ind in layers:\n",
    "    #print(f\"freezing layer {unet.layers[ind].name}\")\n",
    "    unet.layers[ind].trainable = False\n",
    "\n",
    "unet.compile(optimizer=\"adam\",\n",
    "             loss=model.tversky_loss,\n",
    "             loss_weights={\"mask\": 0.3, \"edge\": 0.7},\n",
    "             metrics={'mask': [model.mean_iou, model.dsc, model.tversky, model.tversky_b, 'acc'], \n",
    "                      'edge': [model.mean_iou, model.dsc, model.tversky, model.tversky_b, 'acc']})\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "layers = [(layer.name, layer.output.shape, layer.trainable) for layer in unet.layers]\n",
    "df = pd.DataFrame(layers, columns=['Layer Name', 'Output Shape', 'Layer Trainable'])\n",
    "\n",
    "print(df.to_string())\n",
    "\n",
    "#unet.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- evaluate the model before train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-06-04T22:14:35.605686Z",
     "iopub.status.busy": "2022-06-04T22:14:35.604449Z",
     "iopub.status.idle": "2022-06-04T22:14:56.812000Z",
     "shell.execute_reply": "2022-06-04T22:14:56.810188Z",
     "shell.execute_reply.started": "2022-06-04T22:14:35.605604Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /usr/local/lib/python3.8/dist-packages/tensorflow/python/util/deprecation.py:620: calling map_fn_v2 (from tensorflow.python.ops.map_fn) with dtype is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use fn_output_signature instead\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-06-04 22:14:39.362895: I tensorflow/stream_executor/cuda/cuda_dnn.cc:377] Loaded cuDNN version 8302\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "229/229 [==============================] - 21s 70ms/step - loss: 0.1982 - mask_loss: 0.1982 - edge_loss: 0.7633 - mask_mean_iou: 0.5314 - mask_dsc: 0.7989 - mask_tversky: 0.8018 - mask_tversky_b: 0.9588 - mask_acc: 0.9740 - edge_mean_iou: 0.0576 - edge_dsc: 0.1677 - edge_tversky: 0.2367 - edge_tversky_b: 0.3794 - edge_acc: 0.7129\n"
     ]
    }
   ],
   "source": [
    "old_database = database\n",
    "database = \"data\"\n",
    "\n",
    "test_img_tile_files = sorted(glob.glob(f'{database}/test/tiles/*.jpg'))\n",
    "test_mask_tile_label_files = sorted(glob.glob(f'{database}/test/mask/tiles/*.jpg'))\n",
    "test_edge_tile_label_files = sorted(glob.glob(f'{database}/test/edge/tiles/*.jpg'))\n",
    "\n",
    "old_test_gen = test_gen\n",
    "test_gen = DataGenerator(test_img_tile_files, test_mask_tile_label_files, test_edge_tile_label_files, _batch_size, skip_empty = False, augment=False)\n",
    "\n",
    "unet.evaluate(test_gen)\n",
    "\n",
    "database = old_database\n",
    "test_gen = old_test_gen"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Train the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-06-04T23:45:21.662958Z",
     "iopub.status.busy": "2022-06-04T23:45:21.662489Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.3440 - mask_loss: 0.1254 - edge_loss: 0.4377 - mask_mean_iou: 0.6880 - mask_dsc: 0.8618 - mask_tversky: 0.8746 - mask_tversky_b: 0.8750 - mask_acc: 0.9430 - edge_mean_iou: 0.2911 - edge_dsc: 0.4896 - edge_tversky: 0.5623 - edge_tversky_b: 0.5626 - edge_acc: 0.9228\n",
      "Epoch 00001: val_loss improved from inf to 0.29064, saving model to models/RBC_BinaryCrossentropy_data2_data2_60_data_data_200_03_07__1654386321.7190397_adv_best.h5\n",
      "125/125 [==============================] - 33s 229ms/step - loss: 0.3440 - mask_loss: 0.1254 - edge_loss: 0.4377 - mask_mean_iou: 0.6880 - mask_dsc: 0.8618 - mask_tversky: 0.8746 - mask_tversky_b: 0.8750 - mask_acc: 0.9430 - edge_mean_iou: 0.2911 - edge_dsc: 0.4896 - edge_tversky: 0.5623 - edge_tversky_b: 0.5626 - edge_acc: 0.9228 - val_loss: 0.2906 - val_mask_loss: 0.0685 - val_edge_loss: 0.3858 - val_mask_mean_iou: 0.7567 - val_mask_dsc: 0.9182 - val_mask_tversky: 0.9315 - val_mask_tversky_b: 0.9611 - val_mask_acc: 0.9579 - val_edge_mean_iou: 0.3221 - val_edge_dsc: 0.5560 - val_edge_tversky: 0.6142 - val_edge_tversky_b: 0.6154 - val_edge_acc: 0.9136\n",
      "Epoch 2/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2701 - mask_loss: 0.0434 - edge_loss: 0.3673 - mask_mean_iou: 0.7196 - mask_dsc: 0.9460 - mask_tversky: 0.9566 - mask_tversky_b: 0.9568 - mask_acc: 0.9591 - edge_mean_iou: 0.3137 - edge_dsc: 0.5539 - edge_tversky: 0.6327 - edge_tversky_b: 0.6330 - edge_acc: 0.9273\n",
      "Epoch 00002: val_loss improved from 0.29064 to 0.28941, saving model to models/RBC_BinaryCrossentropy_data2_data2_60_data_data_200_03_07__1654386321.7190397_adv_best.h5\n",
      "125/125 [==============================] - 28s 217ms/step - loss: 0.2701 - mask_loss: 0.0434 - edge_loss: 0.3673 - mask_mean_iou: 0.7196 - mask_dsc: 0.9460 - mask_tversky: 0.9566 - mask_tversky_b: 0.9568 - mask_acc: 0.9591 - edge_mean_iou: 0.3137 - edge_dsc: 0.5539 - edge_tversky: 0.6327 - edge_tversky_b: 0.6330 - edge_acc: 0.9273 - val_loss: 0.2894 - val_mask_loss: 0.0367 - val_edge_loss: 0.3977 - val_mask_mean_iou: 0.7595 - val_mask_dsc: 0.9534 - val_mask_tversky: 0.9633 - val_mask_tversky_b: 0.9634 - val_mask_acc: 0.9592 - val_edge_mean_iou: 0.3192 - val_edge_dsc: 0.5395 - val_edge_tversky: 0.6023 - val_edge_tversky_b: 0.6025 - val_edge_acc: 0.9104\n",
      "Epoch 3/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2764 - mask_loss: 0.0470 - edge_loss: 0.3747 - mask_mean_iou: 0.7197 - mask_dsc: 0.9434 - mask_tversky: 0.9530 - mask_tversky_b: 0.9531 - mask_acc: 0.9578 - edge_mean_iou: 0.3085 - edge_dsc: 0.5448 - edge_tversky: 0.6253 - edge_tversky_b: 0.6255 - edge_acc: 0.9225\n",
      "Epoch 00003: val_loss improved from 0.28941 to 0.28387, saving model to models/RBC_BinaryCrossentropy_data2_data2_60_data_data_200_03_07__1654386321.7190397_adv_best.h5\n",
      "125/125 [==============================] - 28s 217ms/step - loss: 0.2764 - mask_loss: 0.0470 - edge_loss: 0.3747 - mask_mean_iou: 0.7197 - mask_dsc: 0.9434 - mask_tversky: 0.9530 - mask_tversky_b: 0.9531 - mask_acc: 0.9578 - edge_mean_iou: 0.3085 - edge_dsc: 0.5448 - edge_tversky: 0.6253 - edge_tversky_b: 0.6255 - edge_acc: 0.9225 - val_loss: 0.2839 - val_mask_loss: 0.0374 - val_edge_loss: 0.3895 - val_mask_mean_iou: 0.7639 - val_mask_dsc: 0.9562 - val_mask_tversky: 0.9626 - val_mask_tversky_b: 0.9628 - val_mask_acc: 0.9618 - val_edge_mean_iou: 0.3270 - val_edge_dsc: 0.5474 - val_edge_tversky: 0.6105 - val_edge_tversky_b: 0.6106 - val_edge_acc: 0.9120\n",
      "Epoch 4/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2700 - mask_loss: 0.0416 - edge_loss: 0.3679 - mask_mean_iou: 0.7190 - mask_dsc: 0.9470 - mask_tversky: 0.9584 - mask_tversky_b: 0.9586 - mask_acc: 0.9598 - edge_mean_iou: 0.3123 - edge_dsc: 0.5521 - edge_tversky: 0.6321 - edge_tversky_b: 0.6325 - edge_acc: 0.9274\n",
      "Epoch 00004: val_loss improved from 0.28387 to 0.28076, saving model to models/RBC_BinaryCrossentropy_data2_data2_60_data_data_200_03_07__1654386321.7190397_adv_best.h5\n",
      "125/125 [==============================] - 27s 211ms/step - loss: 0.2700 - mask_loss: 0.0416 - edge_loss: 0.3679 - mask_mean_iou: 0.7190 - mask_dsc: 0.9470 - mask_tversky: 0.9584 - mask_tversky_b: 0.9586 - mask_acc: 0.9598 - edge_mean_iou: 0.3123 - edge_dsc: 0.5521 - edge_tversky: 0.6321 - edge_tversky_b: 0.6325 - edge_acc: 0.9274 - val_loss: 0.2808 - val_mask_loss: 0.0372 - val_edge_loss: 0.3852 - val_mask_mean_iou: 0.7600 - val_mask_dsc: 0.9541 - val_mask_tversky: 0.9628 - val_mask_tversky_b: 0.9629 - val_mask_acc: 0.9598 - val_edge_mean_iou: 0.3165 - val_edge_dsc: 0.5335 - val_edge_tversky: 0.6148 - val_edge_tversky_b: 0.6151 - val_edge_acc: 0.8993\n",
      "Epoch 5/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2706 - mask_loss: 0.0415 - edge_loss: 0.3688 - mask_mean_iou: 0.7246 - mask_dsc: 0.9486 - mask_tversky: 0.9585 - mask_tversky_b: 0.9587 - mask_acc: 0.9608 - edge_mean_iou: 0.3147 - edge_dsc: 0.5522 - edge_tversky: 0.6312 - edge_tversky_b: 0.6315 - edge_acc: 0.9284\n",
      "Epoch 00005: val_loss did not improve from 0.28076\n",
      "125/125 [==============================] - 27s 207ms/step - loss: 0.2706 - mask_loss: 0.0415 - edge_loss: 0.3688 - mask_mean_iou: 0.7246 - mask_dsc: 0.9486 - mask_tversky: 0.9585 - mask_tversky_b: 0.9587 - mask_acc: 0.9608 - edge_mean_iou: 0.3147 - edge_dsc: 0.5522 - edge_tversky: 0.6312 - edge_tversky_b: 0.6315 - edge_acc: 0.9284 - val_loss: 0.2880 - val_mask_loss: 0.0368 - val_edge_loss: 0.3956 - val_mask_mean_iou: 0.7626 - val_mask_dsc: 0.9551 - val_mask_tversky: 0.9632 - val_mask_tversky_b: 0.9634 - val_mask_acc: 0.9604 - val_edge_mean_iou: 0.3079 - val_edge_dsc: 0.5219 - val_edge_tversky: 0.6044 - val_edge_tversky_b: 0.6046 - val_edge_acc: 0.8953\n",
      "Epoch 6/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2754 - mask_loss: 0.0421 - edge_loss: 0.3754 - mask_mean_iou: 0.7278 - mask_dsc: 0.9473 - mask_tversky: 0.9579 - mask_tversky_b: 0.9580 - mask_acc: 0.9594 - edge_mean_iou: 0.3081 - edge_dsc: 0.5428 - edge_tversky: 0.6246 - edge_tversky_b: 0.6249 - edge_acc: 0.9237\n",
      "Epoch 00006: val_loss did not improve from 0.28076\n",
      "125/125 [==============================] - 28s 219ms/step - loss: 0.2754 - mask_loss: 0.0421 - edge_loss: 0.3754 - mask_mean_iou: 0.7278 - mask_dsc: 0.9473 - mask_tversky: 0.9579 - mask_tversky_b: 0.9580 - mask_acc: 0.9594 - edge_mean_iou: 0.3081 - edge_dsc: 0.5428 - edge_tversky: 0.6246 - edge_tversky_b: 0.6249 - edge_acc: 0.9237 - val_loss: 0.2907 - val_mask_loss: 0.0457 - val_edge_loss: 0.3957 - val_mask_mean_iou: 0.7534 - val_mask_dsc: 0.9479 - val_mask_tversky: 0.9543 - val_mask_tversky_b: 0.9544 - val_mask_acc: 0.9551 - val_edge_mean_iou: 0.3232 - val_edge_dsc: 0.5436 - val_edge_tversky: 0.6043 - val_edge_tversky_b: 0.6045 - val_edge_acc: 0.9126\n",
      "Epoch 7/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2682 - mask_loss: 0.0389 - edge_loss: 0.3664 - mask_mean_iou: 0.7335 - mask_dsc: 0.9497 - mask_tversky: 0.9611 - mask_tversky_b: 0.9612 - mask_acc: 0.9614 - edge_mean_iou: 0.3170 - edge_dsc: 0.5508 - edge_tversky: 0.6336 - edge_tversky_b: 0.6338 - edge_acc: 0.9235\n",
      "Epoch 00007: val_loss did not improve from 0.28076\n",
      "125/125 [==============================] - 27s 209ms/step - loss: 0.2682 - mask_loss: 0.0389 - edge_loss: 0.3664 - mask_mean_iou: 0.7335 - mask_dsc: 0.9497 - mask_tversky: 0.9611 - mask_tversky_b: 0.9612 - mask_acc: 0.9614 - edge_mean_iou: 0.3170 - edge_dsc: 0.5508 - edge_tversky: 0.6336 - edge_tversky_b: 0.6338 - edge_acc: 0.9235 - val_loss: 0.2883 - val_mask_loss: 0.0400 - val_edge_loss: 0.3947 - val_mask_mean_iou: 0.7585 - val_mask_dsc: 0.9529 - val_mask_tversky: 0.9600 - val_mask_tversky_b: 0.9601 - val_mask_acc: 0.9590 - val_edge_mean_iou: 0.3245 - val_edge_dsc: 0.5435 - val_edge_tversky: 0.6053 - val_edge_tversky_b: 0.6055 - val_edge_acc: 0.9122\n",
      "Epoch 8/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2698 - mask_loss: 0.0407 - edge_loss: 0.3679 - mask_mean_iou: 0.7310 - mask_dsc: 0.9483 - mask_tversky: 0.9593 - mask_tversky_b: 0.9593 - mask_acc: 0.9607 - edge_mean_iou: 0.3164 - edge_dsc: 0.5508 - edge_tversky: 0.6321 - edge_tversky_b: 0.6323 - edge_acc: 0.9244\n",
      "Epoch 00008: val_loss did not improve from 0.28076\n",
      "125/125 [==============================] - 27s 207ms/step - loss: 0.2698 - mask_loss: 0.0407 - edge_loss: 0.3679 - mask_mean_iou: 0.7310 - mask_dsc: 0.9483 - mask_tversky: 0.9593 - mask_tversky_b: 0.9593 - mask_acc: 0.9607 - edge_mean_iou: 0.3164 - edge_dsc: 0.5508 - edge_tversky: 0.6321 - edge_tversky_b: 0.6323 - edge_acc: 0.9244 - val_loss: 0.2915 - val_mask_loss: 0.0379 - val_edge_loss: 0.4002 - val_mask_mean_iou: 0.7557 - val_mask_dsc: 0.9514 - val_mask_tversky: 0.9621 - val_mask_tversky_b: 0.9622 - val_mask_acc: 0.9571 - val_edge_mean_iou: 0.3180 - val_edge_dsc: 0.5378 - val_edge_tversky: 0.5998 - val_edge_tversky_b: 0.6000 - val_edge_acc: 0.9106\n",
      "Epoch 9/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2656 - mask_loss: 0.0390 - edge_loss: 0.3627 - mask_mean_iou: 0.7342 - mask_dsc: 0.9511 - mask_tversky: 0.9610 - mask_tversky_b: 0.9611 - mask_acc: 0.9622 - edge_mean_iou: 0.3209 - edge_dsc: 0.5570 - edge_tversky: 0.6373 - edge_tversky_b: 0.6375 - edge_acc: 0.9264\n",
      "Epoch 00009: val_loss did not improve from 0.28076\n",
      "125/125 [==============================] - 27s 210ms/step - loss: 0.2656 - mask_loss: 0.0390 - edge_loss: 0.3627 - mask_mean_iou: 0.7342 - mask_dsc: 0.9511 - mask_tversky: 0.9610 - mask_tversky_b: 0.9611 - mask_acc: 0.9622 - edge_mean_iou: 0.3209 - edge_dsc: 0.5570 - edge_tversky: 0.6373 - edge_tversky_b: 0.6375 - edge_acc: 0.9264 - val_loss: 0.2916 - val_mask_loss: 0.0384 - val_edge_loss: 0.4001 - val_mask_mean_iou: 0.7604 - val_mask_dsc: 0.9534 - val_mask_tversky: 0.9616 - val_mask_tversky_b: 0.9617 - val_mask_acc: 0.9594 - val_edge_mean_iou: 0.3261 - val_edge_dsc: 0.5462 - val_edge_tversky: 0.5999 - val_edge_tversky_b: 0.6001 - val_edge_acc: 0.9162\n",
      "Epoch 10/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2660 - mask_loss: 0.0368 - edge_loss: 0.3642 - mask_mean_iou: 0.7430 - mask_dsc: 0.9531 - mask_tversky: 0.9632 - mask_tversky_b: 0.9633 - mask_acc: 0.9630 - edge_mean_iou: 0.3216 - edge_dsc: 0.5548 - edge_tversky: 0.6358 - edge_tversky_b: 0.6361 - edge_acc: 0.9224\n",
      "Epoch 00010: val_loss did not improve from 0.28076\n",
      "125/125 [==============================] - 28s 213ms/step - loss: 0.2660 - mask_loss: 0.0368 - edge_loss: 0.3642 - mask_mean_iou: 0.7430 - mask_dsc: 0.9531 - mask_tversky: 0.9632 - mask_tversky_b: 0.9633 - mask_acc: 0.9630 - edge_mean_iou: 0.3216 - edge_dsc: 0.5548 - edge_tversky: 0.6358 - edge_tversky_b: 0.6361 - edge_acc: 0.9224 - val_loss: 0.2882 - val_mask_loss: 0.0381 - val_edge_loss: 0.3954 - val_mask_mean_iou: 0.7564 - val_mask_dsc: 0.9513 - val_mask_tversky: 0.9619 - val_mask_tversky_b: 0.9620 - val_mask_acc: 0.9571 - val_edge_mean_iou: 0.3163 - val_edge_dsc: 0.5340 - val_edge_tversky: 0.6046 - val_edge_tversky_b: 0.6048 - val_edge_acc: 0.9055\n",
      "Epoch 11/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2726 - mask_loss: 0.0441 - edge_loss: 0.3704 - mask_mean_iou: 0.7115 - mask_dsc: 0.9438 - mask_tversky: 0.9559 - mask_tversky_b: 0.9560 - mask_acc: 0.9579 - edge_mean_iou: 0.3083 - edge_dsc: 0.5474 - edge_tversky: 0.6296 - edge_tversky_b: 0.6296 - edge_acc: 0.9263\n",
      "Epoch 00011: val_loss did not improve from 0.28076\n",
      "125/125 [==============================] - 27s 209ms/step - loss: 0.2726 - mask_loss: 0.0441 - edge_loss: 0.3704 - mask_mean_iou: 0.7115 - mask_dsc: 0.9438 - mask_tversky: 0.9559 - mask_tversky_b: 0.9560 - mask_acc: 0.9579 - edge_mean_iou: 0.3083 - edge_dsc: 0.5474 - edge_tversky: 0.6296 - edge_tversky_b: 0.6296 - edge_acc: 0.9263 - val_loss: 0.2809 - val_mask_loss: 0.0375 - val_edge_loss: 0.3852 - val_mask_mean_iou: 0.7660 - val_mask_dsc: 0.9567 - val_mask_tversky: 0.9625 - val_mask_tversky_b: 0.9626 - val_mask_acc: 0.9627 - val_edge_mean_iou: 0.3238 - val_edge_dsc: 0.5418 - val_edge_tversky: 0.6148 - val_edge_tversky_b: 0.6150 - val_edge_acc: 0.9061\n",
      "Epoch 12/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2705 - mask_loss: 0.0406 - edge_loss: 0.3691 - mask_mean_iou: 0.7234 - mask_dsc: 0.9490 - mask_tversky: 0.9594 - mask_tversky_b: 0.9595 - mask_acc: 0.9620 - edge_mean_iou: 0.3140 - edge_dsc: 0.5506 - edge_tversky: 0.6309 - edge_tversky_b: 0.6312 - edge_acc: 0.9248\n",
      "Epoch 00012: val_loss did not improve from 0.28076\n",
      "125/125 [==============================] - 27s 207ms/step - loss: 0.2705 - mask_loss: 0.0406 - edge_loss: 0.3691 - mask_mean_iou: 0.7234 - mask_dsc: 0.9490 - mask_tversky: 0.9594 - mask_tversky_b: 0.9595 - mask_acc: 0.9620 - edge_mean_iou: 0.3140 - edge_dsc: 0.5506 - edge_tversky: 0.6309 - edge_tversky_b: 0.6312 - edge_acc: 0.9248 - val_loss: 0.2857 - val_mask_loss: 0.0357 - val_edge_loss: 0.3929 - val_mask_mean_iou: 0.7653 - val_mask_dsc: 0.9573 - val_mask_tversky: 0.9643 - val_mask_tversky_b: 0.9644 - val_mask_acc: 0.9628 - val_edge_mean_iou: 0.3308 - val_edge_dsc: 0.5504 - val_edge_tversky: 0.6071 - val_edge_tversky_b: 0.6073 - val_edge_acc: 0.9158\n",
      "Epoch 13/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2637 - mask_loss: 0.0385 - edge_loss: 0.3602 - mask_mean_iou: 0.7261 - mask_dsc: 0.9504 - mask_tversky: 0.9615 - mask_tversky_b: 0.9616 - mask_acc: 0.9620 - edge_mean_iou: 0.3205 - edge_dsc: 0.5582 - edge_tversky: 0.6398 - edge_tversky_b: 0.6400 - edge_acc: 0.9273\n",
      "Epoch 00013: val_loss did not improve from 0.28076\n",
      "125/125 [==============================] - 27s 210ms/step - loss: 0.2637 - mask_loss: 0.0385 - edge_loss: 0.3602 - mask_mean_iou: 0.7261 - mask_dsc: 0.9504 - mask_tversky: 0.9615 - mask_tversky_b: 0.9616 - mask_acc: 0.9620 - edge_mean_iou: 0.3205 - edge_dsc: 0.5582 - edge_tversky: 0.6398 - edge_tversky_b: 0.6400 - edge_acc: 0.9273 - val_loss: 0.2816 - val_mask_loss: 0.0357 - val_edge_loss: 0.3870 - val_mask_mean_iou: 0.7682 - val_mask_dsc: 0.9586 - val_mask_tversky: 0.9643 - val_mask_tversky_b: 0.9644 - val_mask_acc: 0.9636 - val_edge_mean_iou: 0.3285 - val_edge_dsc: 0.5497 - val_edge_tversky: 0.6130 - val_edge_tversky_b: 0.6132 - val_edge_acc: 0.9127\n",
      "Epoch 14/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2624 - mask_loss: 0.0363 - edge_loss: 0.3593 - mask_mean_iou: 0.7293 - mask_dsc: 0.9530 - mask_tversky: 0.9637 - mask_tversky_b: 0.9638 - mask_acc: 0.9638 - edge_mean_iou: 0.3198 - edge_dsc: 0.5594 - edge_tversky: 0.6407 - edge_tversky_b: 0.6409 - edge_acc: 0.9265\n",
      "Epoch 00014: val_loss did not improve from 0.28076\n",
      "125/125 [==============================] - 27s 212ms/step - loss: 0.2624 - mask_loss: 0.0363 - edge_loss: 0.3593 - mask_mean_iou: 0.7293 - mask_dsc: 0.9530 - mask_tversky: 0.9637 - mask_tversky_b: 0.9638 - mask_acc: 0.9638 - edge_mean_iou: 0.3198 - edge_dsc: 0.5594 - edge_tversky: 0.6407 - edge_tversky_b: 0.6409 - edge_acc: 0.9265 - val_loss: 0.2839 - val_mask_loss: 0.0363 - val_edge_loss: 0.3901 - val_mask_mean_iou: 0.7579 - val_mask_dsc: 0.9526 - val_mask_tversky: 0.9637 - val_mask_tversky_b: 0.9638 - val_mask_acc: 0.9577 - val_edge_mean_iou: 0.3187 - val_edge_dsc: 0.5368 - val_edge_tversky: 0.6099 - val_edge_tversky_b: 0.6100 - val_edge_acc: 0.9045\n",
      "Epoch 15/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2690 - mask_loss: 0.0428 - edge_loss: 0.3659 - mask_mean_iou: 0.7443 - mask_dsc: 0.9470 - mask_tversky: 0.9572 - mask_tversky_b: 0.9573 - mask_acc: 0.9588 - edge_mean_iou: 0.3243 - edge_dsc: 0.5536 - edge_tversky: 0.6341 - edge_tversky_b: 0.6343 - edge_acc: 0.9238\n",
      "Epoch 00015: val_loss did not improve from 0.28076\n",
      "125/125 [==============================] - 28s 218ms/step - loss: 0.2690 - mask_loss: 0.0428 - edge_loss: 0.3659 - mask_mean_iou: 0.7443 - mask_dsc: 0.9470 - mask_tversky: 0.9572 - mask_tversky_b: 0.9573 - mask_acc: 0.9588 - edge_mean_iou: 0.3243 - edge_dsc: 0.5536 - edge_tversky: 0.6341 - edge_tversky_b: 0.6343 - edge_acc: 0.9238 - val_loss: 0.2841 - val_mask_loss: 0.0357 - val_edge_loss: 0.3906 - val_mask_mean_iou: 0.7659 - val_mask_dsc: 0.9577 - val_mask_tversky: 0.9643 - val_mask_tversky_b: 0.9644 - val_mask_acc: 0.9628 - val_edge_mean_iou: 0.3284 - val_edge_dsc: 0.5497 - val_edge_tversky: 0.6094 - val_edge_tversky_b: 0.6095 - val_edge_acc: 0.9142\n",
      "Epoch 16/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2644 - mask_loss: 0.0375 - edge_loss: 0.3617 - mask_mean_iou: 0.7362 - mask_dsc: 0.9526 - mask_tversky: 0.9625 - mask_tversky_b: 0.9626 - mask_acc: 0.9641 - edge_mean_iou: 0.3219 - edge_dsc: 0.5588 - edge_tversky: 0.6383 - edge_tversky_b: 0.6385 - edge_acc: 0.9267\n",
      "Epoch 00016: val_loss did not improve from 0.28076\n",
      "125/125 [==============================] - 27s 209ms/step - loss: 0.2644 - mask_loss: 0.0375 - edge_loss: 0.3617 - mask_mean_iou: 0.7362 - mask_dsc: 0.9526 - mask_tversky: 0.9625 - mask_tversky_b: 0.9626 - mask_acc: 0.9641 - edge_mean_iou: 0.3219 - edge_dsc: 0.5588 - edge_tversky: 0.6383 - edge_tversky_b: 0.6385 - edge_acc: 0.9267 - val_loss: 0.2862 - val_mask_loss: 0.0389 - val_edge_loss: 0.3921 - val_mask_mean_iou: 0.7619 - val_mask_dsc: 0.9549 - val_mask_tversky: 0.9611 - val_mask_tversky_b: 0.9612 - val_mask_acc: 0.9609 - val_edge_mean_iou: 0.3278 - val_edge_dsc: 0.5454 - val_edge_tversky: 0.6079 - val_edge_tversky_b: 0.6080 - val_edge_acc: 0.9121\n",
      "Epoch 17/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2687 - mask_loss: 0.0417 - edge_loss: 0.3660 - mask_mean_iou: 0.7316 - mask_dsc: 0.9476 - mask_tversky: 0.9583 - mask_tversky_b: 0.9584 - mask_acc: 0.9607 - edge_mean_iou: 0.3200 - edge_dsc: 0.5528 - edge_tversky: 0.6340 - edge_tversky_b: 0.6342 - edge_acc: 0.9262\n",
      "Epoch 00017: val_loss improved from 0.28076 to 0.27957, saving model to models/RBC_BinaryCrossentropy_data2_data2_60_data_data_200_03_07__1654386321.7190397_adv_best.h5\n",
      "125/125 [==============================] - 28s 221ms/step - loss: 0.2687 - mask_loss: 0.0417 - edge_loss: 0.3660 - mask_mean_iou: 0.7316 - mask_dsc: 0.9476 - mask_tversky: 0.9583 - mask_tversky_b: 0.9584 - mask_acc: 0.9607 - edge_mean_iou: 0.3200 - edge_dsc: 0.5528 - edge_tversky: 0.6340 - edge_tversky_b: 0.6342 - edge_acc: 0.9262 - val_loss: 0.2796 - val_mask_loss: 0.0333 - val_edge_loss: 0.3851 - val_mask_mean_iou: 0.7653 - val_mask_dsc: 0.9567 - val_mask_tversky: 0.9667 - val_mask_tversky_b: 0.9668 - val_mask_acc: 0.9619 - val_edge_mean_iou: 0.3312 - val_edge_dsc: 0.5521 - val_edge_tversky: 0.6149 - val_edge_tversky_b: 0.6150 - val_edge_acc: 0.9136\n",
      "Epoch 18/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2627 - mask_loss: 0.0381 - edge_loss: 0.3590 - mask_mean_iou: 0.7370 - mask_dsc: 0.9512 - mask_tversky: 0.9619 - mask_tversky_b: 0.9621 - mask_acc: 0.9629 - edge_mean_iou: 0.3257 - edge_dsc: 0.5597 - edge_tversky: 0.6410 - edge_tversky_b: 0.6412 - edge_acc: 0.9258\n",
      "Epoch 00018: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 28s 215ms/step - loss: 0.2627 - mask_loss: 0.0381 - edge_loss: 0.3590 - mask_mean_iou: 0.7370 - mask_dsc: 0.9512 - mask_tversky: 0.9619 - mask_tversky_b: 0.9621 - mask_acc: 0.9629 - edge_mean_iou: 0.3257 - edge_dsc: 0.5597 - edge_tversky: 0.6410 - edge_tversky_b: 0.6412 - edge_acc: 0.9258 - val_loss: 0.2903 - val_mask_loss: 0.0381 - val_edge_loss: 0.3984 - val_mask_mean_iou: 0.7531 - val_mask_dsc: 0.9502 - val_mask_tversky: 0.9619 - val_mask_tversky_b: 0.9619 - val_mask_acc: 0.9560 - val_edge_mean_iou: 0.3147 - val_edge_dsc: 0.5339 - val_edge_tversky: 0.6016 - val_edge_tversky_b: 0.6018 - val_edge_acc: 0.9067\n",
      "Epoch 19/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2634 - mask_loss: 0.0397 - edge_loss: 0.3593 - mask_mean_iou: 0.7324 - mask_dsc: 0.9499 - mask_tversky: 0.9603 - mask_tversky_b: 0.9604 - mask_acc: 0.9623 - edge_mean_iou: 0.3226 - edge_dsc: 0.5589 - edge_tversky: 0.6407 - edge_tversky_b: 0.6408 - edge_acc: 0.9258\n",
      "Epoch 00019: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 27s 207ms/step - loss: 0.2634 - mask_loss: 0.0397 - edge_loss: 0.3593 - mask_mean_iou: 0.7324 - mask_dsc: 0.9499 - mask_tversky: 0.9603 - mask_tversky_b: 0.9604 - mask_acc: 0.9623 - edge_mean_iou: 0.3226 - edge_dsc: 0.5589 - edge_tversky: 0.6407 - edge_tversky_b: 0.6408 - edge_acc: 0.9258 - val_loss: 0.2934 - val_mask_loss: 0.0808 - val_edge_loss: 0.3844 - val_mask_mean_iou: 0.7557 - val_mask_dsc: 0.8877 - val_mask_tversky: 0.9192 - val_mask_tversky_b: 0.9209 - val_mask_acc: 0.9020 - val_edge_mean_iou: 0.3325 - val_edge_dsc: 0.5543 - val_edge_tversky: 0.6156 - val_edge_tversky_b: 0.6157 - val_edge_acc: 0.9148\n",
      "Epoch 20/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2676 - mask_loss: 0.0430 - edge_loss: 0.3639 - mask_mean_iou: 0.7170 - mask_dsc: 0.9471 - mask_tversky: 0.9570 - mask_tversky_b: 0.9571 - mask_acc: 0.9609 - edge_mean_iou: 0.3149 - edge_dsc: 0.5565 - edge_tversky: 0.6361 - edge_tversky_b: 0.6362 - edge_acc: 0.9280\n",
      "Epoch 00020: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 28s 214ms/step - loss: 0.2676 - mask_loss: 0.0430 - edge_loss: 0.3639 - mask_mean_iou: 0.7170 - mask_dsc: 0.9471 - mask_tversky: 0.9570 - mask_tversky_b: 0.9571 - mask_acc: 0.9609 - edge_mean_iou: 0.3149 - edge_dsc: 0.5565 - edge_tversky: 0.6361 - edge_tversky_b: 0.6362 - edge_acc: 0.9280 - val_loss: 0.2860 - val_mask_loss: 0.0359 - val_edge_loss: 0.3931 - val_mask_mean_iou: 0.7615 - val_mask_dsc: 0.9546 - val_mask_tversky: 0.9641 - val_mask_tversky_b: 0.9642 - val_mask_acc: 0.9601 - val_edge_mean_iou: 0.3285 - val_edge_dsc: 0.5466 - val_edge_tversky: 0.6069 - val_edge_tversky_b: 0.6070 - val_edge_acc: 0.9135\n",
      "Epoch 21/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2677 - mask_loss: 0.0391 - edge_loss: 0.3657 - mask_mean_iou: 0.7456 - mask_dsc: 0.9514 - mask_tversky: 0.9609 - mask_tversky_b: 0.9610 - mask_acc: 0.9630 - edge_mean_iou: 0.3233 - edge_dsc: 0.5533 - edge_tversky: 0.6343 - edge_tversky_b: 0.6346 - edge_acc: 0.9260\n",
      "Epoch 00021: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 27s 208ms/step - loss: 0.2677 - mask_loss: 0.0391 - edge_loss: 0.3657 - mask_mean_iou: 0.7456 - mask_dsc: 0.9514 - mask_tversky: 0.9609 - mask_tversky_b: 0.9610 - mask_acc: 0.9630 - edge_mean_iou: 0.3233 - edge_dsc: 0.5533 - edge_tversky: 0.6343 - edge_tversky_b: 0.6346 - edge_acc: 0.9260 - val_loss: 0.2816 - val_mask_loss: 0.0377 - val_edge_loss: 0.3861 - val_mask_mean_iou: 0.7576 - val_mask_dsc: 0.9527 - val_mask_tversky: 0.9623 - val_mask_tversky_b: 0.9624 - val_mask_acc: 0.9581 - val_edge_mean_iou: 0.3229 - val_edge_dsc: 0.5433 - val_edge_tversky: 0.6139 - val_edge_tversky_b: 0.6140 - val_edge_acc: 0.9077\n",
      "Epoch 22/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2684 - mask_loss: 0.0375 - edge_loss: 0.3674 - mask_mean_iou: 0.7361 - mask_dsc: 0.9523 - mask_tversky: 0.9625 - mask_tversky_b: 0.9626 - mask_acc: 0.9632 - edge_mean_iou: 0.3189 - edge_dsc: 0.5533 - edge_tversky: 0.6326 - edge_tversky_b: 0.6328 - edge_acc: 0.9261\n",
      "Epoch 00022: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 28s 214ms/step - loss: 0.2684 - mask_loss: 0.0375 - edge_loss: 0.3674 - mask_mean_iou: 0.7361 - mask_dsc: 0.9523 - mask_tversky: 0.9625 - mask_tversky_b: 0.9626 - mask_acc: 0.9632 - edge_mean_iou: 0.3189 - edge_dsc: 0.5533 - edge_tversky: 0.6326 - edge_tversky_b: 0.6328 - edge_acc: 0.9261 - val_loss: 0.3012 - val_mask_loss: 0.0379 - val_edge_loss: 0.4141 - val_mask_mean_iou: 0.7549 - val_mask_dsc: 0.9502 - val_mask_tversky: 0.9621 - val_mask_tversky_b: 0.9622 - val_mask_acc: 0.9560 - val_edge_mean_iou: 0.3114 - val_edge_dsc: 0.5300 - val_edge_tversky: 0.5859 - val_edge_tversky_b: 0.5861 - val_edge_acc: 0.9114\n",
      "Epoch 23/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2670 - mask_loss: 0.0398 - edge_loss: 0.3644 - mask_mean_iou: 0.7188 - mask_dsc: 0.9496 - mask_tversky: 0.9602 - mask_tversky_b: 0.9604 - mask_acc: 0.9622 - edge_mean_iou: 0.3140 - edge_dsc: 0.5533 - edge_tversky: 0.6356 - edge_tversky_b: 0.6358 - edge_acc: 0.9263\n",
      "Epoch 00023: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 28s 213ms/step - loss: 0.2670 - mask_loss: 0.0398 - edge_loss: 0.3644 - mask_mean_iou: 0.7188 - mask_dsc: 0.9496 - mask_tversky: 0.9602 - mask_tversky_b: 0.9604 - mask_acc: 0.9622 - edge_mean_iou: 0.3140 - edge_dsc: 0.5533 - edge_tversky: 0.6356 - edge_tversky_b: 0.6358 - edge_acc: 0.9263 - val_loss: 0.2957 - val_mask_loss: 0.0565 - val_edge_loss: 0.3983 - val_mask_mean_iou: 0.7009 - val_mask_dsc: 0.9123 - val_mask_tversky: 0.9435 - val_mask_tversky_b: 0.9438 - val_mask_acc: 0.9183 - val_edge_mean_iou: 0.3061 - val_edge_dsc: 0.5217 - val_edge_tversky: 0.6017 - val_edge_tversky_b: 0.6019 - val_edge_acc: 0.8969\n",
      "Epoch 24/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2652 - mask_loss: 0.0400 - edge_loss: 0.3617 - mask_mean_iou: 0.7318 - mask_dsc: 0.9501 - mask_tversky: 0.9600 - mask_tversky_b: 0.9601 - mask_acc: 0.9620 - edge_mean_iou: 0.3171 - edge_dsc: 0.5561 - edge_tversky: 0.6383 - edge_tversky_b: 0.6384 - edge_acc: 0.9257\n",
      "Epoch 00024: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 27s 207ms/step - loss: 0.2652 - mask_loss: 0.0400 - edge_loss: 0.3617 - mask_mean_iou: 0.7318 - mask_dsc: 0.9501 - mask_tversky: 0.9600 - mask_tversky_b: 0.9601 - mask_acc: 0.9620 - edge_mean_iou: 0.3171 - edge_dsc: 0.5561 - edge_tversky: 0.6383 - edge_tversky_b: 0.6384 - edge_acc: 0.9257 - val_loss: 0.2929 - val_mask_loss: 0.0385 - val_edge_loss: 0.4020 - val_mask_mean_iou: 0.7556 - val_mask_dsc: 0.9505 - val_mask_tversky: 0.9615 - val_mask_tversky_b: 0.9616 - val_mask_acc: 0.9563 - val_edge_mean_iou: 0.3170 - val_edge_dsc: 0.5356 - val_edge_tversky: 0.5980 - val_edge_tversky_b: 0.5982 - val_edge_acc: 0.9098\n",
      "Epoch 25/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2685 - mask_loss: 0.0382 - edge_loss: 0.3672 - mask_mean_iou: 0.7350 - mask_dsc: 0.9519 - mask_tversky: 0.9618 - mask_tversky_b: 0.9619 - mask_acc: 0.9628 - edge_mean_iou: 0.3188 - edge_dsc: 0.5511 - edge_tversky: 0.6328 - edge_tversky_b: 0.6330 - edge_acc: 0.9249\n",
      "Epoch 00025: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 28s 213ms/step - loss: 0.2685 - mask_loss: 0.0382 - edge_loss: 0.3672 - mask_mean_iou: 0.7350 - mask_dsc: 0.9519 - mask_tversky: 0.9618 - mask_tversky_b: 0.9619 - mask_acc: 0.9628 - edge_mean_iou: 0.3188 - edge_dsc: 0.5511 - edge_tversky: 0.6328 - edge_tversky_b: 0.6330 - edge_acc: 0.9249 - val_loss: 0.2948 - val_mask_loss: 0.0457 - val_edge_loss: 0.4015 - val_mask_mean_iou: 0.7581 - val_mask_dsc: 0.9520 - val_mask_tversky: 0.9543 - val_mask_tversky_b: 0.9544 - val_mask_acc: 0.9588 - val_edge_mean_iou: 0.3212 - val_edge_dsc: 0.5421 - val_edge_tversky: 0.5985 - val_edge_tversky_b: 0.5987 - val_edge_acc: 0.9140\n",
      "Epoch 26/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2657 - mask_loss: 0.0397 - edge_loss: 0.3626 - mask_mean_iou: 0.7241 - mask_dsc: 0.9516 - mask_tversky: 0.9603 - mask_tversky_b: 0.9604 - mask_acc: 0.9633 - edge_mean_iou: 0.3178 - edge_dsc: 0.5587 - edge_tversky: 0.6374 - edge_tversky_b: 0.6375 - edge_acc: 0.9284\n",
      "Epoch 00026: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 28s 219ms/step - loss: 0.2657 - mask_loss: 0.0397 - edge_loss: 0.3626 - mask_mean_iou: 0.7241 - mask_dsc: 0.9516 - mask_tversky: 0.9603 - mask_tversky_b: 0.9604 - mask_acc: 0.9633 - edge_mean_iou: 0.3178 - edge_dsc: 0.5587 - edge_tversky: 0.6374 - edge_tversky_b: 0.6375 - edge_acc: 0.9284 - val_loss: 0.2875 - val_mask_loss: 0.0388 - val_edge_loss: 0.3940 - val_mask_mean_iou: 0.7652 - val_mask_dsc: 0.9569 - val_mask_tversky: 0.9612 - val_mask_tversky_b: 0.9613 - val_mask_acc: 0.9626 - val_edge_mean_iou: 0.3249 - val_edge_dsc: 0.5445 - val_edge_tversky: 0.6060 - val_edge_tversky_b: 0.6061 - val_edge_acc: 0.9124\n",
      "Epoch 27/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2650 - mask_loss: 0.0393 - edge_loss: 0.3617 - mask_mean_iou: 0.7490 - mask_dsc: 0.9506 - mask_tversky: 0.9607 - mask_tversky_b: 0.9608 - mask_acc: 0.9607 - edge_mean_iou: 0.3262 - edge_dsc: 0.5572 - edge_tversky: 0.6383 - edge_tversky_b: 0.6385 - edge_acc: 0.9239\n",
      "Epoch 00027: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 28s 214ms/step - loss: 0.2650 - mask_loss: 0.0393 - edge_loss: 0.3617 - mask_mean_iou: 0.7490 - mask_dsc: 0.9506 - mask_tversky: 0.9607 - mask_tversky_b: 0.9608 - mask_acc: 0.9607 - edge_mean_iou: 0.3262 - edge_dsc: 0.5572 - edge_tversky: 0.6383 - edge_tversky_b: 0.6385 - edge_acc: 0.9239 - val_loss: 0.2894 - val_mask_loss: 0.0364 - val_edge_loss: 0.3978 - val_mask_mean_iou: 0.7644 - val_mask_dsc: 0.9559 - val_mask_tversky: 0.9636 - val_mask_tversky_b: 0.9637 - val_mask_acc: 0.9617 - val_edge_mean_iou: 0.3234 - val_edge_dsc: 0.5431 - val_edge_tversky: 0.6022 - val_edge_tversky_b: 0.6023 - val_edge_acc: 0.9130\n",
      "Epoch 28/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2620 - mask_loss: 0.0399 - edge_loss: 0.3571 - mask_mean_iou: 0.7422 - mask_dsc: 0.9504 - mask_tversky: 0.9601 - mask_tversky_b: 0.9602 - mask_acc: 0.9614 - edge_mean_iou: 0.3267 - edge_dsc: 0.5615 - edge_tversky: 0.6429 - edge_tversky_b: 0.6430 - edge_acc: 0.9263\n",
      "Epoch 00028: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 28s 219ms/step - loss: 0.2620 - mask_loss: 0.0399 - edge_loss: 0.3571 - mask_mean_iou: 0.7422 - mask_dsc: 0.9504 - mask_tversky: 0.9601 - mask_tversky_b: 0.9602 - mask_acc: 0.9614 - edge_mean_iou: 0.3267 - edge_dsc: 0.5615 - edge_tversky: 0.6429 - edge_tversky_b: 0.6430 - edge_acc: 0.9263 - val_loss: 0.2822 - val_mask_loss: 0.0390 - val_edge_loss: 0.3865 - val_mask_mean_iou: 0.7662 - val_mask_dsc: 0.9570 - val_mask_tversky: 0.9610 - val_mask_tversky_b: 0.9611 - val_mask_acc: 0.9627 - val_edge_mean_iou: 0.3311 - val_edge_dsc: 0.5539 - val_edge_tversky: 0.6135 - val_edge_tversky_b: 0.6137 - val_edge_acc: 0.9152\n",
      "Epoch 29/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2632 - mask_loss: 0.0386 - edge_loss: 0.3595 - mask_mean_iou: 0.7443 - mask_dsc: 0.9521 - mask_tversky: 0.9614 - mask_tversky_b: 0.9616 - mask_acc: 0.9633 - edge_mean_iou: 0.3252 - edge_dsc: 0.5583 - edge_tversky: 0.6405 - edge_tversky_b: 0.6407 - edge_acc: 0.9258\n",
      "Epoch 00029: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 27s 209ms/step - loss: 0.2632 - mask_loss: 0.0386 - edge_loss: 0.3595 - mask_mean_iou: 0.7443 - mask_dsc: 0.9521 - mask_tversky: 0.9614 - mask_tversky_b: 0.9616 - mask_acc: 0.9633 - edge_mean_iou: 0.3252 - edge_dsc: 0.5583 - edge_tversky: 0.6405 - edge_tversky_b: 0.6407 - edge_acc: 0.9258 - val_loss: 0.3098 - val_mask_loss: 0.0721 - val_edge_loss: 0.4117 - val_mask_mean_iou: 0.6683 - val_mask_dsc: 0.8879 - val_mask_tversky: 0.9279 - val_mask_tversky_b: 0.9279 - val_mask_acc: 0.8933 - val_edge_mean_iou: 0.2969 - val_edge_dsc: 0.5120 - val_edge_tversky: 0.5883 - val_edge_tversky_b: 0.5885 - val_edge_acc: 0.8964\n",
      "Epoch 30/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2631 - mask_loss: 0.0367 - edge_loss: 0.3601 - mask_mean_iou: 0.7275 - mask_dsc: 0.9528 - mask_tversky: 0.9633 - mask_tversky_b: 0.9634 - mask_acc: 0.9634 - edge_mean_iou: 0.3176 - edge_dsc: 0.5584 - edge_tversky: 0.6399 - edge_tversky_b: 0.6401 - edge_acc: 0.9268\n",
      "Epoch 00030: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 28s 215ms/step - loss: 0.2631 - mask_loss: 0.0367 - edge_loss: 0.3601 - mask_mean_iou: 0.7275 - mask_dsc: 0.9528 - mask_tversky: 0.9633 - mask_tversky_b: 0.9634 - mask_acc: 0.9634 - edge_mean_iou: 0.3176 - edge_dsc: 0.5584 - edge_tversky: 0.6399 - edge_tversky_b: 0.6401 - edge_acc: 0.9268 - val_loss: 0.2823 - val_mask_loss: 0.0365 - val_edge_loss: 0.3876 - val_mask_mean_iou: 0.7627 - val_mask_dsc: 0.9555 - val_mask_tversky: 0.9635 - val_mask_tversky_b: 0.9635 - val_mask_acc: 0.9610 - val_edge_mean_iou: 0.3271 - val_edge_dsc: 0.5467 - val_edge_tversky: 0.6124 - val_edge_tversky_b: 0.6125 - val_edge_acc: 0.9110\n",
      "Epoch 31/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2624 - mask_loss: 0.0395 - edge_loss: 0.3579 - mask_mean_iou: 0.7453 - mask_dsc: 0.9504 - mask_tversky: 0.9605 - mask_tversky_b: 0.9606 - mask_acc: 0.9609 - edge_mean_iou: 0.3276 - edge_dsc: 0.5598 - edge_tversky: 0.6421 - edge_tversky_b: 0.6423 - edge_acc: 0.9247\n",
      "Epoch 00031: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 27s 211ms/step - loss: 0.2624 - mask_loss: 0.0395 - edge_loss: 0.3579 - mask_mean_iou: 0.7453 - mask_dsc: 0.9504 - mask_tversky: 0.9605 - mask_tversky_b: 0.9606 - mask_acc: 0.9609 - edge_mean_iou: 0.3276 - edge_dsc: 0.5598 - edge_tversky: 0.6421 - edge_tversky_b: 0.6423 - edge_acc: 0.9247 - val_loss: 0.2802 - val_mask_loss: 0.0354 - val_edge_loss: 0.3852 - val_mask_mean_iou: 0.7602 - val_mask_dsc: 0.9529 - val_mask_tversky: 0.9646 - val_mask_tversky_b: 0.9647 - val_mask_acc: 0.9586 - val_edge_mean_iou: 0.3296 - val_edge_dsc: 0.5505 - val_edge_tversky: 0.6148 - val_edge_tversky_b: 0.6150 - val_edge_acc: 0.9125\n",
      "Epoch 32/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2665 - mask_loss: 0.0379 - edge_loss: 0.3645 - mask_mean_iou: 0.7353 - mask_dsc: 0.9526 - mask_tversky: 0.9621 - mask_tversky_b: 0.9622 - mask_acc: 0.9628 - edge_mean_iou: 0.3188 - edge_dsc: 0.5543 - edge_tversky: 0.6355 - edge_tversky_b: 0.6356 - edge_acc: 0.9254\n",
      "Epoch 00032: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 27s 210ms/step - loss: 0.2665 - mask_loss: 0.0379 - edge_loss: 0.3645 - mask_mean_iou: 0.7353 - mask_dsc: 0.9526 - mask_tversky: 0.9621 - mask_tversky_b: 0.9622 - mask_acc: 0.9628 - edge_mean_iou: 0.3188 - edge_dsc: 0.5543 - edge_tversky: 0.6355 - edge_tversky_b: 0.6356 - edge_acc: 0.9254 - val_loss: 0.2895 - val_mask_loss: 0.0371 - val_edge_loss: 0.3977 - val_mask_mean_iou: 0.7520 - val_mask_dsc: 0.9471 - val_mask_tversky: 0.9629 - val_mask_tversky_b: 0.9630 - val_mask_acc: 0.9537 - val_edge_mean_iou: 0.3179 - val_edge_dsc: 0.5370 - val_edge_tversky: 0.6023 - val_edge_tversky_b: 0.6024 - val_edge_acc: 0.9090\n",
      "Epoch 33/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2667 - mask_loss: 0.0375 - edge_loss: 0.3649 - mask_mean_iou: 0.7418 - mask_dsc: 0.9524 - mask_tversky: 0.9625 - mask_tversky_b: 0.9626 - mask_acc: 0.9627 - edge_mean_iou: 0.3198 - edge_dsc: 0.5544 - edge_tversky: 0.6351 - edge_tversky_b: 0.6352 - edge_acc: 0.9252\n",
      "Epoch 00033: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 28s 219ms/step - loss: 0.2667 - mask_loss: 0.0375 - edge_loss: 0.3649 - mask_mean_iou: 0.7418 - mask_dsc: 0.9524 - mask_tversky: 0.9625 - mask_tversky_b: 0.9626 - mask_acc: 0.9627 - edge_mean_iou: 0.3198 - edge_dsc: 0.5544 - edge_tversky: 0.6351 - edge_tversky_b: 0.6352 - edge_acc: 0.9252 - val_loss: 0.2866 - val_mask_loss: 0.0362 - val_edge_loss: 0.3939 - val_mask_mean_iou: 0.7629 - val_mask_dsc: 0.9550 - val_mask_tversky: 0.9638 - val_mask_tversky_b: 0.9638 - val_mask_acc: 0.9607 - val_edge_mean_iou: 0.3237 - val_edge_dsc: 0.5425 - val_edge_tversky: 0.6061 - val_edge_tversky_b: 0.6062 - val_edge_acc: 0.9109\n",
      "Epoch 34/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2706 - mask_loss: 0.0457 - edge_loss: 0.3669 - mask_mean_iou: 0.7161 - mask_dsc: 0.9446 - mask_tversky: 0.9543 - mask_tversky_b: 0.9544 - mask_acc: 0.9576 - edge_mean_iou: 0.3124 - edge_dsc: 0.5523 - edge_tversky: 0.6331 - edge_tversky_b: 0.6331 - edge_acc: 0.9257\n",
      "Epoch 00034: val_loss did not improve from 0.27957\n",
      "125/125 [==============================] - 28s 214ms/step - loss: 0.2706 - mask_loss: 0.0457 - edge_loss: 0.3669 - mask_mean_iou: 0.7161 - mask_dsc: 0.9446 - mask_tversky: 0.9543 - mask_tversky_b: 0.9544 - mask_acc: 0.9576 - edge_mean_iou: 0.3124 - edge_dsc: 0.5523 - edge_tversky: 0.6331 - edge_tversky_b: 0.6331 - edge_acc: 0.9257 - val_loss: 0.2847 - val_mask_loss: 0.0357 - val_edge_loss: 0.3913 - val_mask_mean_iou: 0.7627 - val_mask_dsc: 0.9554 - val_mask_tversky: 0.9643 - val_mask_tversky_b: 0.9644 - val_mask_acc: 0.9607 - val_edge_mean_iou: 0.3264 - val_edge_dsc: 0.5465 - val_edge_tversky: 0.6087 - val_edge_tversky_b: 0.6088 - val_edge_acc: 0.9125\n",
      "Epoch 35/200\n",
      "125/125 [==============================] - ETA: 0s - loss: 0.2665 - mask_loss: 0.0442 - edge_loss: 0.3618 - mask_mean_iou: 0.7384 - mask_dsc: 0.9451 - mask_tversky: 0.9558 - mask_tversky_b: 0.9558 - mask_acc: 0.9595 - edge_mean_iou: 0.3278 - edge_dsc: 0.5567 - edge_tversky: 0.6382 - edge_tversky_b: 0.6383 - edge_acc: 0.9271\n",
      "Epoch 00035: val_loss improved from 0.27957 to 0.27383, saving model to models/RBC_BinaryCrossentropy_data2_data2_60_data_data_200_03_07__1654386321.7190397_adv_best.h5\n",
      "125/125 [==============================] - 28s 218ms/step - loss: 0.2665 - mask_loss: 0.0442 - edge_loss: 0.3618 - mask_mean_iou: 0.7384 - mask_dsc: 0.9451 - mask_tversky: 0.9558 - mask_tversky_b: 0.9558 - mask_acc: 0.9595 - edge_mean_iou: 0.3278 - edge_dsc: 0.5567 - edge_tversky: 0.6382 - edge_tversky_b: 0.6383 - edge_acc: 0.9271 - val_loss: 0.2738 - val_mask_loss: 0.0337 - val_edge_loss: 0.3767 - val_mask_mean_iou: 0.7703 - val_mask_dsc: 0.9598 - val_mask_tversky: 0.9663 - val_mask_tversky_b: 0.9664 - val_mask_acc: 0.9649 - val_edge_mean_iou: 0.3344 - val_edge_dsc: 0.5552 - val_edge_tversky: 0.6233 - val_edge_tversky_b: 0.6234 - val_edge_acc: 0.9119\n",
      "Epoch 36/200\n",
      "109/125 [=========================>....] - ETA: 2s - loss: 0.2677 - mask_loss: 0.0376 - edge_loss: 0.3663 - mask_mean_iou: 0.7414 - mask_dsc: 0.9520 - mask_tversky: 0.9624 - mask_tversky_b: 0.9624 - mask_acc: 0.9622 - edge_mean_iou: 0.3197 - edge_dsc: 0.5512 - edge_tversky: 0.6337 - edge_tversky_b: 0.6338 - edge_acc: 0.9228"
     ]
    }
   ],
   "source": [
    "if Train:\n",
    "    _model_name_new = _model_name + \"_data_data_200_03_07__\" + str(time.time()) + \"_adv\"\n",
    "    try:\n",
    "        os.remove(f\"models/{_model_name}_best.h5\")\n",
    "        os.remove(f\"models/{_model_name}_all.h5\")\n",
    "    except:\n",
    "        pass\n",
    "    \n",
    "    history = unet.fit(train_gen, \n",
    "            epochs=_epochs, \n",
    "            validation_data=test_gen, \n",
    "            steps_per_epoch=train_steps,\n",
    "            max_queue_size=2*_workers,\n",
    "            use_multiprocessing=True,\n",
    "            workers=_workers,\n",
    "            verbose=1,\n",
    "            callbacks=model.get_callbacks(_model_name_new))\n",
    "    \n",
    "    unet.save(f'./models/{_model_name_new}',save_format='tf')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- save the training log "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#save the training log into a file\n",
    "import json\n",
    "my_dict = {'a':11,'b':22,'c':33}\n",
    "with open(f\"models/{_model_name}_log.json\", \"a+\") as fp:\n",
    "    json.dump(history.history,fp)\n",
    "\n",
    "# Opening JSON file\n",
    "with open(f\"models/{_model_name}_log.json\") as fp:\n",
    "    # returns JSON object as\n",
    "    # a dictionary\n",
    "    loaded_hist = json.load(fp)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- evaluate model after training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "old_database = database\n",
    "database = \"data\"\n",
    "\n",
    "test_img_tile_files = sorted(glob.glob(f'{database}/test/tiles/*.jpg'))\n",
    "test_mask_tile_label_files = sorted(glob.glob(f'{database}/test/mask/tiles/*.jpg'))\n",
    "test_edge_tile_label_files = sorted(glob.glob(f'{database}/test/edge/tiles/*.jpg'))\n",
    "\n",
    "old_test_gen = test_gen\n",
    "test_gen = DataGenerator(test_img_tile_files, test_mask_tile_label_files, test_edge_tile_label_files, _batch_size, skip_empty = False, augment=False)\n",
    "\n",
    "unet.evaluate(test_gen)\n",
    "\n",
    "database = old_database\n",
    "test_gen = old_test_gen"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "images, (masks, edges) = train_gen.__getitem__(165)\n",
    "\n",
    "#data.plot_ime(images, masks, edges)\n",
    "\n",
    "prediction = unet.predict(images)\n",
    "\n",
    "#prediction = np.squeeze(prediction)\n",
    "\n",
    "data.plot_ime(images, prediction[0], prediction[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-06-04T23:10:32.194369Z",
     "iopub.status.busy": "2022-06-04T23:10:32.194112Z",
     "iopub.status.idle": "2022-06-04T23:10:32.402104Z",
     "shell.execute_reply": "2022-06-04T23:10:32.400399Z",
     "shell.execute_reply.started": "2022-06-04T23:10:32.194342Z"
    }
   },
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'img' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m/tmp/ipykernel_107/2617079263.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0max\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfig\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd_subplot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m4\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfig_num\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0max\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mset_title\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"original image\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 6\u001b[0;31m \u001b[0max\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mimshow\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mimg\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      7\u001b[0m \u001b[0mfig_num\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfig_num\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[0max\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfig\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd_subplot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m4\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfig_num\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'img' is not defined"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1200x800 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(15, 10), dpi=80)\n",
    "fig.subplots_adjust(hspace=0.3, wspace=0.3)\n",
    "fig_num = 1\n",
    "ax = fig.add_subplot(2, 4, fig_num)\n",
    "ax.set_title(\"original image\")\n",
    "ax.imshow(img[0])\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(2, 4, fig_num)\n",
    "ax.set_title(\"original mask\")\n",
    "ax.imshow(mask)\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(2, 4, fig_num)\n",
    "ax.set_title(\" original edge mask\")\n",
    "ax.imshow(edge)\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(2, 4, fig_num)\n",
    "ax.set_title(\" substraction\")\n",
    "ax.imshow((mask - edge) > 0)\n",
    "\n",
    "\n",
    "fig_num = fig_num + 2\n",
    "ax = fig.add_subplot(2, 4, fig_num)\n",
    "ax.set_title(\"out_mask\")\n",
    "ax.imshow(prediction[0])\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(2, 4, fig_num)\n",
    "ax.set_title(\"out_edge\")\n",
    "ax.imshow(prediction[1])\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(2, 4, fig_num)\n",
    "ax.set_title(\"substraction\")\n",
    "ax.imshow((prediction[0] - prediction[1]) > 0)\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- predict full image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "images, masks, edges = model.predictFullImage(unet,\n",
    "                        data.load_data_na(sorted(glob.glob(f'data/test/*.jpg'))[:20] + glob.glob(f'*.jpg'), RGB=True, clahe=True),\n",
    "                        padding=100,\n",
    "                        input_size=188,\n",
    "                        output_size=100,\n",
    "                        normalize_output = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "data.plot_ime(images,\n",
    "              masks,\n",
    "              edges,\n",
    "              figSize=(25,15))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## model training data plot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "hist = history.history\n",
    "\n",
    "lines = 4\n",
    "columns = 3\n",
    "\n",
    "fig = plt.figure(figsize=(20, 20), dpi=80)\n",
    "fig.subplots_adjust(hspace=0.3, wspace=0.3)\n",
    "fig_num = 1\n",
    "ax = fig.add_subplot(lines, columns, fig_num)\n",
    "ax.set_title(\"mask accuracy\")\n",
    "ax.plot(hist['mask_acc'], label=\"Training\")\n",
    "ax.plot(hist['val_mask_acc'], label=\"Validation\")\n",
    "ax.legend()\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(lines, columns, fig_num)\n",
    "ax.set_title(\"edge mask accuracy\")\n",
    "ax.plot(hist['edge_acc'], label=\"Training\")\n",
    "ax.plot(hist['val_edge_acc'], label=\"Validation\")\n",
    "ax.legend()\n",
    "fig_num = fig_num + 2\n",
    "ax = fig.add_subplot(lines, columns, fig_num)\n",
    "ax.set_title(\"mask loss\")\n",
    "ax.plot(hist['mask_loss'], label=\"Training\")\n",
    "ax.plot(hist['val_mask_loss'], label=\"Validation\")\n",
    "ax.legend()\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(lines, columns, fig_num)\n",
    "ax.set_title(\"edge loss\")\n",
    "ax.plot(hist['edge_loss'], label=\"Training\")\n",
    "ax.plot(hist['val_edge_loss'], label=\"Validation\")\n",
    "ax.legend()\n",
    "\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(lines, columns, fig_num)\n",
    "ax.set_title(\"global loss\")\n",
    "ax.plot(hist['loss'], label=\"Training\")\n",
    "ax.plot(hist['val_loss'], label=\"Validation\")\n",
    "ax.legend()\n",
    "\n",
    "\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(lines, columns, fig_num)\n",
    "ax.set_title(\"mask metrics\")\n",
    "ax.plot(hist['mask_dsc'], label=\"dsc\")\n",
    "ax.plot(hist['mask_mean_iou'], label=\"mean iou\")\n",
    "ax.plot(hist['mask_tversky'], label=\"tversky\")\n",
    "ax.plot(hist['mask_acc'], label=\"accuracy\")\n",
    "\n",
    "ax.set_title(\"mask metrics\")\n",
    "ax.plot(hist['val_mask_dsc'], label=\"val_dsc\")\n",
    "ax.plot(hist['val_mask_mean_iou'], label=\"val_mean iou\")\n",
    "ax.plot(hist['val_mask_tversky'], label=\"val_tversky\")\n",
    "ax.plot(hist['val_mask_acc'], label=\"val_accuracy\")\n",
    "\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(lines, columns, fig_num)\n",
    "ax.set_title(\"edge metrics\")\n",
    "ax.plot(hist['edge_dsc'], label=\"dsc\")\n",
    "ax.plot(hist['edge_mean_iou'], label=\"mean iou\")\n",
    "ax.plot(hist['edge_tversky'], label=\"tversky\")\n",
    "ax.plot(hist['edge_acc'], label=\"accuracy\")\n",
    "\n",
    "ax.set_title(\"edge metrics\")\n",
    "ax.plot(hist['val_edge_dsc'], label=\"val_dsc\")\n",
    "ax.plot(hist['val_edge_mean_iou'], label=\"val_mean iou\")\n",
    "ax.plot(hist['val_edge_tversky'], label=\"val_tversky\")\n",
    "ax.plot(hist['val_edge_acc'], label=\"val_accuracy\")\n",
    "ax.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## For Debugging\n",
    "- show in between layers output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'pool1' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m/home/zaxo7/WorkSpace/DEEP LEARNING/MY-U-NET/test.ipynb Cell 22'\u001b[0m in \u001b[0;36m<cell line: 3>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      <a href='vscode-notebook-cell:/home/zaxo7/WorkSpace/DEEP%20LEARNING/MY-U-NET/test.ipynb#ch0000022?line=0'>1</a>\u001b[0m ind \u001b[39m=\u001b[39m np\u001b[39m.\u001b[39mrandom\u001b[39m.\u001b[39mrandint(\u001b[39m0\u001b[39m, \u001b[39m32\u001b[39m, \u001b[39m1\u001b[39m)[\u001b[39m0\u001b[39m]\n\u001b[0;32m----> <a href='vscode-notebook-cell:/home/zaxo7/WorkSpace/DEEP%20LEARNING/MY-U-NET/test.ipynb#ch0000022?line=2'>3</a>\u001b[0m _pool1 \u001b[39m=\u001b[39m pool1[:\u001b[39m-\u001b[39m\u001b[39m1\u001b[39m,:\u001b[39m-\u001b[39m\u001b[39m1\u001b[39m,ind]\n\u001b[1;32m      <a href='vscode-notebook-cell:/home/zaxo7/WorkSpace/DEEP%20LEARNING/MY-U-NET/test.ipynb#ch0000022?line=3'>4</a>\u001b[0m _pool2 \u001b[39m=\u001b[39m pool2[:\u001b[39m-\u001b[39m\u001b[39m1\u001b[39m,:\u001b[39m-\u001b[39m\u001b[39m1\u001b[39m,ind]\n\u001b[1;32m      <a href='vscode-notebook-cell:/home/zaxo7/WorkSpace/DEEP%20LEARNING/MY-U-NET/test.ipynb#ch0000022?line=4'>5</a>\u001b[0m _pool3 \u001b[39m=\u001b[39m pool3[:\u001b[39m-\u001b[39m\u001b[39m1\u001b[39m,:\u001b[39m-\u001b[39m\u001b[39m1\u001b[39m,ind]\n",
      "\u001b[0;31mNameError\u001b[0m: name 'pool1' is not defined"
     ]
    }
   ],
   "source": [
    "ind = np.random.randint(0, 32, 1)[0]\n",
    "\n",
    "_pool1 = pool1[:-1,:-1,ind]\n",
    "_pool2 = pool2[:-1,:-1,ind]\n",
    "_pool3 = pool3[:-1,:-1,ind]\n",
    "_down4 = down4[:-1,:-1,ind]\n",
    "_up1 = up1[:-1,:-1,ind]\n",
    "_up2 = up2[:-1,:-1,ind]\n",
    "_up3 = up3[:-1,:-1,ind]\n",
    "\n",
    "fig = plt.figure(figsize=(15, 10), dpi=80)\n",
    "fig.subplots_adjust(hspace=0.3, wspace=0.3)\n",
    "fig_num = 1\n",
    "ax = fig.add_subplot(3, 3, fig_num)\n",
    "ax.set_title(\"out_mask\")\n",
    "ax.imshow(out_mask)\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(3, 3, fig_num)\n",
    "ax.set_title(\"out_edge\")\n",
    "ax.imshow(out_edge)\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(3, 3, fig_num)\n",
    "ax.set_title(\"pool1\")\n",
    "ax.imshow(_pool1)\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(3, 3, fig_num)\n",
    "ax.set_title(\"pool2\")\n",
    "ax.imshow(_pool2)\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(3, 3, fig_num)\n",
    "ax.set_title(\"pool3\")\n",
    "ax.imshow(_pool3)\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(3, 3, fig_num)\n",
    "ax.set_title(\"down4\")\n",
    "ax.imshow(_down4)\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(3, 3, fig_num)\n",
    "ax.set_title(\"up1\")\n",
    "ax.imshow(_up1)\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(3, 3, fig_num)\n",
    "ax.set_title(\"up2\")\n",
    "ax.imshow(_up2)\n",
    "fig_num = fig_num + 1\n",
    "ax = fig.add_subplot(3, 3, fig_num)\n",
    "ax.set_title(\"up3\")\n",
    "ax.imshow(_up3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- function to plot and array of images of shape (width, height, n_images)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def picshow(img, title):\n",
    "    num = img.shape[2]\n",
    "    imgs_per_line = 4\n",
    "    ax = num//imgs_per_line + 1\n",
    "    ay = imgs_per_line\n",
    "    fig =plt.figure(figsize=(30, num//imgs_per_line * 10))\n",
    "    fig.subplots_adjust(hspace=0.1, wspace=0.1)\n",
    "    for i in range(1,num + 1):\n",
    "        sub = fig.add_subplot(ax,ay,i)\n",
    "        sub.set_title(f\"{title} {i}\")\n",
    "        sub.imshow(img[:-1,:-1,i - 1])\n",
    "    plt.show()\n",
    "    \n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "picshow(pool1, 'pool1 ')\n",
    "picshow(pool2, 'pool2 ')\n",
    "picshow(pool3, 'pool3 ')\n",
    "picshow(down4, 'down4 ')\n",
    "picshow(up1, 'up1 ')\n",
    "picshow(up2, 'up2 ')\n",
    "picshow(up3, 'up3 ')"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "31f2aee4e71d21fbe5cf8b01ff0e069b9275f58929596ceb00d14d90e3e16cd6"
  },
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}