Improve README

README.md

@@ -3,6 +3,7 @@ SDFEst is a package for pose, scale, and shape estimation using discretized sign
 
 [Installation](#installation) | [Dataset Preparation](#dataset-preparation)  | [Paper Reproduction](#paper-reproduction) | [Code Structure](#code-structure) | [Development](#development) | [Citation](#citation) | [Docs](https://roym899.github.io/sdfest/)
 
+![Animation of differentiable renderer.](resources/sdfest_overview.gif)
 
 ## Installation
 
@@ -34,7 +35,7 @@ pip install -r requirements.txt -e .
 You need to install Detectron2 manually to run the pipeline with automatic instance segmentation.
 
 Follow the [detectron2 installation guide](https://detectron2.readthedocs.io/en/latest/tutorials/install.html) from there.
-Tested with detectron2 0.5 + torch 1.9.0.
+Tested with detectron2 0.6 + torch 1.12.0 (and various older versions).
 
 ## Dataset Preparation
 See below for expected folder structure for each dataset.
@@ -71,14 +72,14 @@ See below for expected folder structure for each dataset.
 
 First, make sure the datasets are in the right format.
 
-### Benchmark Results
-Depending on which dataset, you have downloaded you can reproduce the results reported in the paper (using the already trained models) by running the script
-```bash
-source reproduce_{shapenet,modelnet,redwood}_experiments.sh
-```
-after that, all results can be found in `./results`.
+### Full paper (training and experiments)
+See the bash script in `reproduction_scripts/reproduce_paper.sh`.
+
+Evaluation code for REAL275 and REDWOOD75 experiments will be integrated in [cpas_toolbox](https://github.com/roym899/pose_and_shape_evaluation) soon.
+
+<sup>Non-cleaned up version of evaluation code can be found in `icaps_eval` branch.</sup>
 
-### Train Models
+### Train Models Only
 To train a network for a specific category you need to first train a per-category VAE, and *afterwards* an initialization network.
 #### VAE
 First we need to convert the ShapeNet meshes to SDFs and optionally filter the dataset. To reproduce the preprocessing of the paper run
@@ -91,7 +92,7 @@ source train_vaes.sh
 ```
 to train the models using the same configuration as used for the paper.
 
-#### Init Network
+#### Initialization Network
 To train the initialization network we used in our paper, run
 ```bash
 source train_init_networks.sh
@@ -113,7 +114,7 @@ Code is structured into 4 sub-packages:
 
 Differentiable rendering of depth image for signed distance fields.
 
-The signed distance field is assumed to be voxelized and it's pose is given by a x, y, z in the camera frame, a quaternion describing its orientation and a scale parameter describing its size. This module provides the derivative with respect to the signed distance values, and the full pose description (position, orientation, scale).
+The signed distance field is assumed to be voxelized and its pose is given by a x, y, z in the camera frame, a quaternion describing its orientation and a scale parameter describing its size. This module provides the derivative with respect to the signed distance values, and the full pose description (position, orientation, scale).
 
 #### Generating compile_commands.json
 <sup>General workflow for PyTorch extensions (only tested for JIT, probably similar otherwise)</sup>