yolact 구현하기

환경

설정

git clone https://github.com/dbolya/yolact.git
cd yolact
pip install opencv-python pillow==6.2.1 pycocotools matplotlib
pip install cython
pip install torch==1.2.0 torchvision==0.4.0    # CUDA10.0で安定

cd external/DCNv2
python setup.py develop

학습된 모델로 추론

python eval.py 
    --trained_model=weights/yolact_base_54_800000.pth  
    --score_threshold=0.15 
    --top_k=15 
    --images=path/to/input/folder:path/to/output/folder

python eval.py 
    --trained_model=weights/yolact_base_54_800000.pth 
    --score_threshold=0.15 
    --top_k=15 
    --video_multiframe=4 
    --video=input_video.mp4:output_video.mp4

원본 데이터로 학습

dataset_name/
├── test
│   ├── test.json
│   └── images/
└── train
    ├── train.json
    └── images/

    test_dataset = dataset_base.copy({
        'name': 'Test Dataset',

        'train_images': '~/dataset_name/train/images/',
        'train_info':   '~/dataset_name/train/train.json',

        'valid_images': '~/dataset_name/test/images/',
        'valid_info':   '~/dataset_name/test/test.json',

        'has_gt': True,
        'class_names': ('class_1', 'class_2', 'class_3', ...),    # クラス名を定義
        'label_map': {'class_id_1': 1, 'class_id_2': 2, ...}    # class idの辞書
    })

    yolact_base_config = coco_base_config.copy({
        'name': 'yolact_base',

        # Dataset stuff
        # 'dataset': coco2017_dataset,    # default
        'dataset': test_dataset,    # Original Dataset

        # 'num_classes': len(coco2017_dataset.class_names) + 1,
        'num_classes': len(test_dataset.class_names) + 1,
    })

python train.py --config=yolact_plus_resnet50_config --save_folder=/path/to/workspace/ --save_interval=1000

python eval.py --trained_model=weights/yolact_base_54_800000.pth