vinhnguyen21/faz-segmentation

Building Unet architecture for Foveal Avascular Zone Extraction

FAZ-Segmentation

Combine Hessian Filter and UNet for Foveal Avascular Zone Extraction

Docker Installation for FLask App

1. Build and run docker on port 2001

$ ./docker-build.sh

If getting error in permission

$ chmod u+x ./docker-build.sh

2. I have already mounted /images to /images of docker, so to test we will prepare image in:

├   ├── images
|       ├── raw
|           ├── 1.tif
|       ├── predict
|           ├── 1.png

• We will put raw image in /images/raw
• In postman:
  • url: http://localhost:2001/faz/predict
  • METHOD: GET
  • Params:
    • Key: id
    • value: name of image such as 1.png
• You will have the prediction of model at /images/predict

Training process

Prepare dataset folder

├── train
|   ├── raw 
|       ├── image1.tif
|       ├── ...
|   ├── mask
|       ├── image1.png
|       ├── ...
├── valid
|   ├── raw 
|       ├── image1.png
|       ├── ...
|   ├── mask
|       ├── image1.png
|       ├── ...
├── test
|   ├── raw 
|       ├── image1.tif
|       ├── ...
|   ├── mask
|       ├── image1.png
|       ├── ...

Setup Environment

Run this script to create a virtual environment and install dependency libraries

1. $conda create -n name_environment python=3.6
2. $conda activate name_environment
3. $pip install -r requirements-2.txt

To train this project, we just run the command

$python train.py

where train_config.json which is located in config folder

We need to adjust the parameter in this json file before training:

• net_type: name of pretrained model you want to train. list of model: efficentnet_b0, efficientnet_b1, efficientnet_b2, efficientnet_b3, efficientnet_b4, efficientnet_b5, Se_resnext50, Se_resnext101, Se_resnet50, se_resnet101, Se_resnet152, Resnet18, Resnet34,Resnet50, Resnet101

• pretrained: boolean, using pretrained weights from ImageNet

• weight_path: Weight path of old trained model

• train_folder : path of raw folder of training dataset example: /home/vinhng/OCTA/preprocess_OCTA/train/raw

• valid_folder : path of raw folder of valid dataset example: /home/vinhng/OCTA/preprocess_OCTA/valid/raw

• test_folder : path of raw folder of valid dataset example: /home/vinhng/OCTA/preprocess_OCTA/test/raw

• classes: number of classes. Default = 1

• model_path: directory which contains trained model

• size: size of input image and mask

• thresh_hold: thresh hold for convert grayscale mask to binary mask

• epoch: number of training epoch

Testing process

download weight of model: 
https://storage.googleapis.com/v-project/Se_resnext50-920eef84.pth

Then move this weight in folder: 
./models

python test.py --path_images --model_type --weight 

• path_images: directory of raw folder in testset (see prepare dataset above)
• model_type: name of pretrained model you want to train. Default: Se_resnext50

List of pretrained model is at training process above

• weight: directory to weight path.