In model/I3D_Pytorch.py line 302 to line 304: > nn.AvgPool3d(kernel_size=(2, 7, 7), stride=1),# (1024, 8, 1, 1) > nn.Dropout3d(dropout_drop_prob), > nn.Conv3d(1024, num_classes, kernel_size=1, stride=1, bias=True),# (400, 8, 1, 1) it seems like the dimensions wrote in the comment should be (1024, 7, 1, 1) and (400, 7, 1, 1) since there's no padding :)
In model/I3D_Pytorch.py line 302 to line 304:
it seems like the dimensions wrote in the comment should be (1024, 7, 1, 1) and (400, 7, 1, 1) since there's no padding :)