check_anchor_order() update

pull/1/head
Glenn Jocher 5 years ago
parent 1f1917ef56
commit fc171e2678

@ -61,8 +61,9 @@ class Model(nn.Module):
# Build strides, anchors # Build strides, anchors
m = self.model[-1] # Detect() m = self.model[-1] # Detect()
m.stride = torch.tensor([64 / x.shape[-2] for x in self.forward(torch.zeros(1, ch, 64, 64))]) # forward m.stride = torch.tensor([128 / x.shape[-2] for x in self.forward(torch.zeros(1, ch, 128, 128))]) # forward
m.anchors /= m.stride.view(-1, 1, 1) m.anchors /= m.stride.view(-1, 1, 1)
check_anchor_order(m)
self.stride = m.stride self.stride = m.stride
# Init weights, biases # Init weights, biases

@ -58,7 +58,8 @@ def check_anchors(dataset, model, thr=4.0, imgsz=640):
print('\nAnalyzing anchors... ', end='') print('\nAnalyzing anchors... ', end='')
m = model.module.model[-1] if hasattr(model, 'module') else model.model[-1] # Detect() m = model.module.model[-1] if hasattr(model, 'module') else model.model[-1] # Detect()
shapes = imgsz * dataset.shapes / dataset.shapes.max(1, keepdims=True) shapes = imgsz * dataset.shapes / dataset.shapes.max(1, keepdims=True)
wh = torch.tensor(np.concatenate([l[:, 3:5] * s for s, l in zip(shapes, dataset.labels)])).float() # wh scale = np.random.uniform(0.9, 1.1, size=(shapes.shape[0], 1)) # augment scale
wh = torch.tensor(np.concatenate([l[:, 3:5] * s for s, l in zip(shapes * scale, dataset.labels)])).float() # wh
def metric(k): # compute metric def metric(k): # compute metric
r = wh[:, None] / k[None] r = wh[:, None] / k[None]
@ -77,12 +78,23 @@ def check_anchors(dataset, model, thr=4.0, imgsz=640):
new_anchors = torch.tensor(new_anchors, device=m.anchors.device).type_as(m.anchors) new_anchors = torch.tensor(new_anchors, device=m.anchors.device).type_as(m.anchors)
m.anchor_grid[:] = new_anchors.clone().view_as(m.anchor_grid) # for inference m.anchor_grid[:] = new_anchors.clone().view_as(m.anchor_grid) # for inference
m.anchors[:] = new_anchors.clone().view_as(m.anchors) / m.stride.to(m.anchors.device).view(-1, 1, 1) # loss m.anchors[:] = new_anchors.clone().view_as(m.anchors) / m.stride.to(m.anchors.device).view(-1, 1, 1) # loss
check_anchor_order(m)
print('New anchors saved to model. Update model *.yaml to use these anchors in the future.') print('New anchors saved to model. Update model *.yaml to use these anchors in the future.')
else: else:
print('Original anchors better than new anchors. Proceeding with original anchors.') print('Original anchors better than new anchors. Proceeding with original anchors.')
print('') # newline print('') # newline
def check_anchor_order(m):
# Check anchor order against stride order for YOLOv5 Detect() module m, and correct if necessary
a = m.anchor_grid.prod(-1).view(-1) # anchor area
da = a[-1] - a[0] # delta a
ds = m.stride[-1] - m.stride[0] # delta s
if da.sign() != ds.sign(): # same order
m.anchors[:] = m.anchors.flip(0)
m.anchor_grid[:] = m.anchor_grid.flip(0)
def check_file(file): def check_file(file):
# Searches for file if not found locally # Searches for file if not found locally
if os.path.isfile(file): if os.path.isfile(file):

Loading…
Cancel
Save