Contextual-Deeplab-V3

python=2.7

Tutorial

Download source code

git clone https://github.com/ecustWallace/cataract.git

Download modules

pip install -r requirements.txt

Download dataset and model checkpoint

Dataset

https://drive.google.com/file/d/13I2xBSyVNAL1YySyZgdHmfr5yl2s3z6G/view?usp=sharing

Make a directory called train in the root directory, and then extract the folder to train.

Dataset Index

https://drive.google.com/file/d/1ByE3X_MGa6gqRKFdH6QydXpKS6MNUHo_/view?usp=sharing

Put the file into train.

Model Checkpoint

https://drive.google.com/file/d/1LpNmBzcgZDrFef-5wnDFdWEVCJ4SeNXY/view?usp=sharing

Make a directory called snapshot in the root directory, and then extract this file into snapshot.

Train

If you want to train from scratch:

python train.py

If train from a pre-trained model:

python train.py --restore-from {your_model}

It can be directly changed in train.py by changing RESTORE_FROM.

Inference

We can make the inference on any folder that contains the images.

python inference_batch.py --data-dir {your_data_dir} --save-dir {your_save_dir} --model {your_model_name}

For example:

python inference_batch.py --data-dir inference/output/train_12+_MP/N3_600_C/ --save-dir