nc: 80
depth_multiple: 1.0
width_multiple: 1.0
anchors: 3
backbone:
[
[-1, 1, Silence, []],
[-1, 1, Conv, [64, 3, 2]],
[-1, 1, Conv, [128, 3, 2]],
[-1, 1, RepNCSPELAN4, [256, 128, 64, 2]],
[-1, 1, ADown, [256]],
[-1, 1, RepNCSPELAN4, [512, 256, 128, 2]],
[-1, 1, ADown, [512]],
[-1, 1, RepNCSPELAN4, [1024, 512, 256, 2]],
[-1, 1, ADown, [1024]],
[-1, 1, RepNCSPELAN4, [1024, 512, 256, 2]],
[1, 1, CBLinear, [[64]]],
[3, 1, CBLinear, [[64, 128]]],
[5, 1, CBLinear, [[64, 128, 256]]],
[7, 1, CBLinear, [[64, 128, 256, 512]]],
[9, 1, CBLinear, [[64, 128, 256, 512, 1024]]],
[0, 1, Conv, [64, 3, 2]],
[[10, 11, 12, 13, 14, -1], 1, CBFuse, [[0, 0, 0, 0, 0]]],
[-1, 1, Conv, [128, 3, 2]],
[[11, 12, 13, 14, -1], 1, CBFuse, [[1, 1, 1, 1]]],
[-1, 1, RepNCSPELAN4, [256, 128, 64, 2]],
[-1, 1, ADown, [256]],
[[12, 13, 14, -1], 1, CBFuse, [[2, 2, 2]]],
[-1, 1, RepNCSPELAN4, [512, 256, 128, 2]],
[-1, 1, ADown, [512]],
[[13, 14, -1], 1, CBFuse, [[3, 3]]],
[-1, 1, RepNCSPELAN4, [1024, 512, 256, 2]],
[-1, 1, ADown, [1024]],
[[14, -1], 1, CBFuse, [[4]]],
[-1, 1, RepNCSPELAN4, [1024, 512, 256, 2]],
]
head:
[
[28, 1, SPPELAN, [512, 256]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 25], 1, Concat, [1]],
[-1, 1, RepNCSPELAN4, [512, 512, 256, 2]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 22], 1, Concat, [1]],
[-1, 1, RepNCSPELAN4, [256, 256, 128, 2]],
[-1, 1, ADown, [256]],
[[-1, 32], 1, Concat, [1]],
[-1, 1, RepNCSPELAN4, [512, 512, 256, 2]],
[-1, 1, ADown, [512]],
[[-1, 29], 1, Concat, [1]],
[-1, 1, RepNCSPELAN4, [512, 1024, 512, 2]],
[[35, 38, 41], 1, DDetect, [nc]],
]