Closed SeongjaeHong closed 3 years ago
bbox_roi_extractor=dict(
type='SingleRoIExtractor',
roi_layer=dict(type='RoIAlign', output_size=7, sampling_ratio=0),
out_channels=256,
featmap_strides=[4, 8, 16, 32]),
You can add the corresponding stride to the featmap_strides.
Feel free to reopen the issue if there is any question.
https://github.com/open-mmlab/mmdetection/blob/dbc6b67446d3898461d0a5d5cef7d112e710db0a/mmdet/models/roi_heads/roi_extractors/single_level_roi_extractor.py#L99-L107
If the goal is to use every feature map, then can I just shorten the above code like this ?
If it's not possible, can you please explain the reason?