siqi9747/cfr-icl-interactive-segmentation

Cascade-Forward Refinement with Iterative Click Loss for Interactive Image Segmentation

CFR-ICL: Cascade-Forward Refinement with Iterative Click Loss for Interactive Image Segmentation

Environment

Training and evaluation environment: Python 3.9, PyTorch 1.13.1, CUDA 11.0. Run the following command to install required packages.

pip3 install -r requirements.txt

You need to configue the paths to the datasets in config.yml before training or testing.

Dataset

A script download_datasets.sh is prepared to download and organize required datasets.

Dataset	Description	Download Link
MS COCO	118k images with 1.2M instances (train)	official site
LVIS v1.0	100k images with 1.2M instances (total)	official site
COCO+LVIS*	99k images with 1.5M instances (train)	original LVIS images + combined annotations
SBD	8498 images with 20172 instances for (train) 2857 images with 6671 instances for (test)	official site
Grab Cut	50 images with one object each (test)	GrabCut.zip (11 MB)
Berkeley	96 images with 100 instances (test)	Berkeley.zip (7 MB)
DAVIS	345 images with one object each (test)	DAVIS.zip (43 MB)
Pascal VOC	1449 images with 3417 instances (test)	official site
COCO_MVal	800 images with 800 instances (test)	COCO_MVal.zip (127 MB)

Demo

An example script to run the demo.

python demo.py --checkpoint=weights/cocolvis_icl_vit_huge.pth --gpu 0

Evaluation

Before evaluation, please download the datasets and models, and then configure the path in config.yml.

Download our model, please download below 3 zipped files and extract before use:

• cocolvis_icl_vit_huge.pth.7z.001
• cocolvis_icl_vit_huge.pth.7z.002
• cocolvis_icl_vit_huge.pth.7z.003

Use the following code to evaluate the huge model.

python scripts/evaluate_model.py NoBRS \
    --gpu=0 \
    --checkpoint=cocolvis_icl_vit_huge.pth \
    --datasets=GrabCut,Berkeley,DAVIS,PascalVOC,SBD \\
    --cf-n=4 \\
    --acf

# cf-n: CFR steps
# acf: adaptive CFR

Training

Before training, please download the MAE pretrained weights (click to download: ViT-Base, ViT-Large, ViT-Huge) and configure the dowloaded path in config.yml

Please also download the pretrained SimpleClick models from here.

Use the following code to train a huge model on C+L:

python train.py models/plainvit_huge448_cocolvis.py \
    --batch-size=32 \
    --ngpus=4

Citation

@article{sun2023cfricl,
      title={CFR-ICL: Cascade-Forward Refinement with Iterative Click Loss for Interactive Image Segmentation}, 
      author={Shoukun Sun and Min Xian and Fei Xu and Tiankai Yao and Luca Capriotti},
      year={2023},
      eprint={2303.05620},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Acknowledgement

Our project is developed based on RITM and SimpleClick