xy and wh losses respectively merged

This commit is contained in:
Glenn Jocher 2019-02-19 22:19:59 +01:00
parent 3eb49be263
commit 0772ebf7c9
1 changed files with 10 additions and 15 deletions

View File

@ -231,18 +231,16 @@ def build_targets(target, anchor_wh, nA, nC, nG):
continue
t = target[b]
# Convert to position relative to box
gx, gy, gw, gh = t[:, 1] * nG, t[:, 2] * nG, t[:, 3] * nG, t[:, 4] * nG
gxy, gwh = t[:, 1:3] * nG, t[:, 3:5] * nG
# Get grid box indices and prevent overflows (i.e. 13.01 on 13 anchors)
gi = torch.clamp(gx.long(), min=0, max=nG - 1)
gj = torch.clamp(gy.long(), min=0, max=nG - 1)
gi, gj = torch.clamp(gxy.long(), min=0, max=nG - 1).t()
# iou of targets-anchors (using wh only)
box1 = t[:, 3:5] * nG
box1 = gwh
box2 = anchor_wh.unsqueeze(1)
inter_area = torch.min(box1, box2).prod(2)
iou = inter_area / (gw * gh + box2.prod(2) - inter_area + 1e-16)
iou = inter_area / (box1.prod(1) + box2.prod(2) - inter_area + 1e-16)
# Select best iou_pred and anchor
iou_best, a = iou.max(0) # best anchor [0-2] for each target
@ -269,17 +267,14 @@ def build_targets(target, anchor_wh, nA, nC, nG):
if iou_best < 0.10:
continue
tc, gx, gy, gwh = t[:, 0].long(), t[:, 1] * nG, t[:, 2] * nG, t[:, 3:5] * nG
tc, gxy, gwh = t[:, 0].long(), t[:, 1:3] * nG, t[:, 3:5] * nG
# Coordinates
txy[b, a, gj, gi, 0] = gx - gi.float()
txy[b, a, gj, gi, 1] = gy - gj.float()
# XY coordinates
txy[b, a, gj, gi] = gxy - gxy.floor()
# Width and height (yolo method)
twh[b, a, gj, gi] = torch.log(gwh / anchor_wh[a])
# Width and height (power method)
# twh[b, a, gj, gi] = torch.sqrt(gwh / anchor_wh[a]) / 2
# Width and height
twh[b, a, gj, gi] = torch.log(gwh / anchor_wh[a]) # yolo method
# twh[b, a, gj, gi] = torch.sqrt(gwh / anchor_wh[a]) / 2 # power method
# One-hot encoding of label
tcls[b, a, gj, gi, tc] = 1