Unverified 提交 bfacfc6b authored 作者: Glenn Jocher's avatar Glenn Jocher 提交者: GitHub

Update cls bias init (#5520)

* Update cls bias init Increased numerical precision. Returns 1.0 probability for single-class datasets now. Addresses https://github.com/ultralytics/yolov5/issues/5357 ```python torch.sigmoid(torch.tensor([math.log(0.6 / (1 - 0.99999))])) Out[19]: tensor([1.0000]) ``` * Update yolo.py
上级 36a4de18
......@@ -201,7 +201,7 @@ class Model(nn.Module):
for mi, s in zip(m.m, m.stride): # from
b = mi.bias.view(m.na, -1) # conv.bias(255) to (3,85)
b.data[:, 4] += math.log(8 / (640 / s) ** 2) # obj (8 objects per 640 image)
b.data[:, 5:] += math.log(0.6 / (m.nc - 0.99)) if cf is None else torch.log(cf / cf.sum()) # cls
b.data[:, 5:] += math.log(0.6 / (m.nc - 0.999999)) if cf is None else torch.log(cf / cf.sum()) # cls
mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True)
def _print_biases(self):
......
Markdown 格式
0%
您添加了 0 到此讨论。请谨慎行事。
请先完成此评论的编辑!
注册 或者 后发表评论