Recall always jumping to zero from first epoch

[Vamshigoud784]

Hi Nikhil,
Thanks for this UNET implementation.
I have followed exact steps in your tutorial and setup and gave all paths correctly.
I have used "train.py" and it is showing empty training data . are there any files me

Terminal log:
python train.py
Train: 4542 - 4542
Test: 1136 - 1136
Train for 379 steps, validate for 95 steps
Epoch 1/10
378/379 [============================>.] - ETA: 0s - loss: 0.0251 - mean_io_u: 1.0000 - recall: 0.0071 - precision: 0.2735
Epoch 00001: saving model to unet.h5
379/379 [==============================] - 273s 719ms/step - loss: 0.0251 - mean_io_u: 1.0000 - recall: 0.0071 - precision: 0.2735 - val_loss: 0.0086 - val_mean_io_u: 1.0000 - val_recall: 0.0000e+00 - val_precision: 0.0000e+00
Epoch 2/10
WARNING:tensorflow:Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least steps_per_epoch * epochs batches (in this case, 3790 batches). You may need to use the repeat() function when building your dataset.

Epoch 00002: saving model to unet.h5
WARNING:tensorflow:Reduce LR on plateau conditioned on metric val_loss which is not available. Available metrics are: lr
WARNING:tensorflow:Early stopping conditioned on metric val_loss which is not available. Available metrics are:
0/379 [..............................] - ETA: 0sTraceback (most recent call last):
File "train.py", line 62, in
callbacks=callbacks
File "/home/anaconda3/envs/UNET/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training.py", line 819, in fit
use_multiprocessing=use_multiprocessing)
File "/home/anaconda3/envs/UNET/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training_v2.py", line 342, in fit
total_epochs=epochs)
File "/home/anaconda3/envs/UNET/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training_v2.py", line 187, in run_one_epoch
aggregator.finalize()
File "/home/anaconda3/envs/UNET/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training_utils.py", line 144, in finalize
raise ValueError('Empty training data.')
ValueError: Empty training data.

[k-sashank]

In the data.py file, under the read_mask() function, there is a line:

x = x/255.0

Please comment that line or remove that line, compile the code and run again.
I am sure this will resolve the issue.

[nikhilroxtomar]

raise ValueError('Empty training data.') ValueError: Empty training data.

Please check the path of the dataset. The above error is because the dataset is empty and I think it is about the validation dataset.

[k-nayak]

Is it possible to have real time segmentation without lag from the camera feed using U-net ?

When i try to feed a video or stream from camera there is huge lag in the feed when compared to just streaming without the model in openCV.

Anyway the lag can be reduced ?