Merge pull request #15 from joubs/14-prepare-release

Prepare release 1.0.0

README.md

@@ -7,8 +7,7 @@ This repository contains an implementation of the CHOWDER method described in
 The code is written in Python3+ and has been tested with version 3.7
 It uses the Pytorch framework.
 
-The package provides a main application that performs both training and evaluation on a given dataset following 
-a certain folder tree structure (as the provided data folder does).
+The package provides a main application that performs both training and evaluation on the Camelyon16 public dataset.
 
 ## Required items
 The repository contains a report (`report.md`) containing:
@@ -37,15 +36,19 @@ mypy chowder                 # Run the static type checking
 flake8                       # Run the lint tool
 ```
 
-** By default, Pytorch is installed with cpuonly version. One can run the application to train on GPU without changing
-the code by installing the CUDA compatible version, please check [this page to do so.](https://pytorch.org/get-started/locally/)
+**By default, Pytorch is installed with cpuonly version. One can run the application to train on GPU without changing
+the code by installing the CUDA compatible version, please check [this page to do so.](https://pytorch.org/get-started/locally/)**
+
 ## Run the application
 
 The main application is registered as an entry point. After installation, it can be run with the following command:
 
 ```chowder_train --data_folder /path/to/the/root/data/folder ```
 
-NOTE: the medical data folder can be pasted directly at the root of the repository which is expected by default. 
+The application expects a path to the `data` folder that was provided alongside the technical test. It relies 
+on its precise tree structure.
+
+**NOTE**: the medical data folder can be pasted directly at the root of the repository which is expected by default. 
 
 The main application is written in `chowder/__main__.py`, one could also launch the application by running this file.
 

report.md

@@ -1,13 +1,13 @@
 # Description of the implemented algorithm.
 
-The CHOWDER algorithm falls under the scope of weak supervised learning. Its primary objective is to predict a binary 
+The CHOWDER algorithm falls under the scope of weak supervised learning. A first objective is to predict a binary 
 label (Healthy VS Cancer) out of a given image. 
 It is designed to operate over WSI images, that have a very high resolution. While the data is annotated at the image 
 level, the network is fed with tiles obtained after sampling the original images. The algorithm exploits tiles 
 descriptors to predict a label at the image level. 
 
-The algorithm can also be used to predict a label at the tile level, a usage could be for example to help a clinician to
-annotate a dataset by suggesting suspicious regions on a WSI image.
+The algorithm can also be used to predict a label at the tile level (segmentation), a usage could be for example to 
+help a clinician to annotate a dataset by suggesting suspicious regions on a WSI image.
 
 
 # Design choices and specifications of the code.

setup.py

@@ -15,7 +15,7 @@
 
 setup(
     name="chowder_impl",
-    version="0.0.0",
+    version="1.0.0",
     description="Implementation of the CHOWDER model for histopathological image analysis.",
     long_description=README,
     long_description_content_type="text/markdown",