ENet-keras
This is an implementation of ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation, ported from ENet-training (lua-torch) to keras.
Requirements
OpenCV, Pillow and scikit-image - These are currently all listed as required. The reason is the project is undergoing a cleanup behind the scenes on this front. Currently OpenCV is used predominantly. However, it's a weird library and I'm attempting to replace it with something else. predict.py is used as a testbed.
hdf5
Installation
Anaconda/Miniconda
conda env create -f environment.yml
pip
pip install -r requirements.txt
pycocotools
Follow the instructions on the repo to install the MS-COCO API.
Preparation
git clone https://github.com/PavlosMelissinos/enet-keras.git
Set up pycocotools
Either set PYTHONPATH accordingly:
export PYTHONPATH=path/to/MS-COCO/API/PythonAPI or
add a symbolic link to the pycocotools directory in the root of the project:
ln -s path/to/MS-COCO/API/PythonAPI/pycocotools .
Prepare data
cd data
./download_mscoco.sh
Prepare pretrained ENet model
TODO
Usage
Predict
python src/predict.py path/to/txt/file/containing/image/paths /path/to/h5/model /path/where/predictions/will/be/saved
Train on MS-COCO
TODO
Remaining tasks
- Open new issue about available image processing libraries.
- download_mscoco.sh should extract the archives to their appropriate locations
- Upload pretrained model
- Finalize prediction.py
- Make data loader multithreaded
- Remove opencv from data loader
- Remove opencv from train.py
- Debug train.py
- Retrain ENet for rgb values