/Chest-Xray-Version2

Using RetinaNet with framework Detectron2 - Pytorch

Primary LanguagePython

Chest-Xray-Version2

Status: done training, creating docker and client webserver

The model detect the abnormalities in chest-Xray image by Detectron2 - Pytorch.

INSTALLATION

  1. Create virtual environment.
conda create -n chestxrayv2 python=3.7
conda activate chestxrayv2
  1. clone this repository.
  2. Install required packages.
pip install -r requirements.txt
  1. Setup Detectron2.

See installation instructions. Or see my instructions.

DATA PREPROCESSING

See the document for understanding how we process the chest-Xray dataset from VinBigdata.

  1. Download the standard and additional data after processing.
bash download_data_standard_add.sh

You can run file streamlit_.py for exploring the dataset in eda/ or nms-wbf-visualize/ .

Note: Need to configure config/streamlit_eda.yaml file.

DOWNLOAD PRETRAIN MODEL:

You can download our model with 5 classes

  1. Download pretrain model with best mAP50 after 5000 epochs.
python experiments-records/download_5_classes_model.py

FOR TRAINING

python train.py

You need configure traininig in config/train.yaml.

FOR EVALUATING

python eval.py

You need configure evaluating in config/inference.yaml.

DEPLOY WEBSITE APPLICATION

streamlit run streamlit_inference.py

You need configure evaluating in config/inference.yaml.