AutoShape integer image-size fix (#10090)

Update common.py We have a division at line 694, and then a multiplication at line 695, so it makes `y*g` not an integer. And since `shape1` will be used at line 697 to ensure the size is divisible by the `stride`, this may lead to different image size. In my experiment, my image is [640, 640], it's divisible by the default stride 32, but I found that the result is changed to [672, 672] after line 697. So the final detection result is slightly different from that directly using the `detect.py` script, which does not call the AutoShape methods. Signed-off-by: janus-zheng <106574221+janus-zheng@users.noreply.github.com> Signed-off-by: janus-zheng <106574221+janus-zheng@users.noreply.github.com> Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>

AutoShape integer image-size fix (#10090)
892c4cd4 · janus-zheng · GitHub · 86decb3c · 892c4cd4
--- a/models/common.py
+++ b/models/common.py
@@ -692,7 +692,7 @@ class AutoShape(nn.Module):
                s = im.shape[:2]  # HWC
                shape0.append(s)  # image shape
                g = max(size) / max(s)  # gain
-                shape1.append([y * g for y in s])
+                shape1.append([int(y * g) for y in s])
                ims[i] = im if im.data.contiguous else np.ascontiguousarray(im)  # update
            shape1 = [make_divisible(x, self.stride) for x in np.array(shape1).max(0)] if self.pt else size  # inf shape
            x = [letterbox(im, shape1, auto=False)[0] for im in ims]  # pad