roi_head question

[javierztl]

Hi, thank you very much for your great work. I have a question that the htc_cbv2_swin_base_patch4_window7_mstrain_400-1400_adamw_20e.py is inherited from htc_without_semantic_swin_fpn.py that illiterately does not use semantic. If I command out the _roihead to remove the _semantic_roiextractor and _semantichead, does it affect the final accuracy. My datasets have mask annotations info but I didn't convert them to .pngs, are they necessary?

[tingtingliangvs]

Hi, Without mask annotation, the performance for object detection may drop a bit. You might need to convert mask annotation into .json format to follow the COCO style.

[javierztl]

My annotation is the coco style. When I start to train, it still required png files, maybe the bug #4 is not fixed well.

[chuxiaojie]

If you wanna train without semantic gt, you may try this:

train_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='LoadAnnotations', with_bbox=True, with_mask=True),
    dict(type='RandomFlip', flip_ratio=0.5),
    dict(
        type='Resize',
        img_scale=[(1600, 400), (1600, 1400)],
        multiscale_mode='range',
        keep_ratio=True),
    dict(type='Normalize', **img_norm_cfg),
    dict(type='Pad', size_divisor=32),
    dict(type='DefaultFormatBundle'),
    dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels', 'gt_masks']),
]

[larsoncs]

I want train without semantic gt, but do not drop mAP, can you help me thks? @tingtingliangvs @chuxiaojie