Closed dan520520520 closed 3 years ago
Please follow the latest official instructions (this) to do the testing. If you still get wrong results, please provide the complete testing log.
@mike112223 filter_widerface_val.py时会遇到中断,把assert np.sum(bbox == gt_box) == 4注释掉了,有影响吗?
Have you solved the problem? I followed the instructions again, and i did not stuck at running filter_widerface_val.py.
Have you solved the problem? I followed the instructions again, and i did not stuck at running filter_widerface_val.py. 解决了,dict(typename='Resize', keep_ratio=True),我之前keep_ratio=False关掉了,打开就可以了。我是按照步骤做,达到了0.923.
不运行filter_widerface_val.py,可以达到0.923吗?
Check #24. If you skip this step, some extremely hard faces will be taken into account in evaluation, which will definitely make the score low.
@mike112223
is filtering bounding box necessary?
I did not see the same thing in retinaface (https://github.com/biubug6/Pytorch_Retinaface/blob/master/test_widerface.py)
@mike112223
is filtering bounding box necessary?
I did not see the same thing in retinaface (https://github.com/biubug6/Pytorch_Retinaface/blob/master/test_widerface.py)
It's necessary.
@mileistone
can you provide the reference?
It's a postprocessing method proposed by us, there is no reference.
@mileistone
I change the backbone to mobilnet0.25 in order to reproduce a better result than https://github.com/biubug6/Pytorch_Retinaface.
However, the result is inferior, for example, the easy set is 84.8 vs 88.67. they use ohem which is inferior than focal loss in theory. Do you have any idea?
tinaface_r50_fpn_widerface.pth测试,按照Issues #25的过程做了一遍,还是下面的结果,做了filter_widerface_val.py以后感觉xml没什么变化? +-------+-------+---------+--------+-------+ | class | gts | dets | recall | ap | +-------+-------+---------+--------+-------+ | face | 31957 | 8420592 | 0.837 | 0.628 | +-------+-------+---------+--------+-------+ | mAP | | | | 0.628 | +-------+-------+---------+--------+-------+