Anchor Number

[Yuanye-F]

I reduce the anchor number from 3 to 2, and there is a problem during training (evaluation):

anchor_boxes[:, :, :, :2] = ((r[:, :, :, :2].sigmoid() * 2. - 0.5) + grid) * stride

RuntimeError: The size of tensor a (2) must match the size of tensor b (3) at non-singleton dimension 3

The model configure is:

[model-configure]
pre_weights=None
classes=7
width=320
height=320
anchor_num=2
anchors=10.54,9.51, 45.60,40.45, 119.62,95.06, 253.71,138.37

[Yuanye-F]

Solve the problem.

In utils/utils.py:

line 300 :
from
return torch.stack((wv, hv), 2).repeat(1,1,3).reshape(h, w, cfg["anchor_num"], -1).to(device)
to
return torch.stack((wv, hv), 2).repeat(1,1,cfg["anchor_num"]).reshape(h, w, cfg["anchor_num"], -1).to(device)

line 326:
from
c = c.repeat(1, 1, 3, 1)
to
c = c.repeat(1, 1, cfg["anchor_num"], 1)