I updated boxnum=2000 in get_random_data and did kmeans on widerface train datasets to get anchors of wideface train dataset.
After 320 epoch widerface training, it is not as good as SSD structure Face detection such as Blazeface,
have you tried widerface train?
if it is possible, can I have your suggestion? Thank You very much.
I updated boxnum=2000 in get_random_data and did kmeans on widerface train datasets to get anchors of wideface train dataset.
After 320 epoch widerface training, it is not as good as SSD structure Face detection such as Blazeface,
have you tried widerface train?
if it is possible, can I have your suggestion? Thank You very much.