Bienqq
/
car-detection-bayes
2
1
Fork 0
Code Issues Pull Requests Releases Wiki Activity

updates

This commit is contained in:
Glenn Jocher 2019-08-03 14:14:10 +02:00
parent 333cf92bb2
commit 90daf8f19c
1 changed files with 18 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

41
models.py
View File

@ -39,42 +39,37 @@ def create_modules(module_defs):
elif module_def['type'] == 'maxpool': elif module_def['type'] == 'maxpool':
kernel_size = int(module_def['size']) kernel_size = int(module_def['size'])
stride = int(module_def['stride']) stride = int(module_def['stride'])
if kernel_size == 2 and stride == 1: if kernel_size == 2 and stride == 1: # yolov3-tiny
modules.add_module('_debug_padding_%d' % i, nn.ZeroPad2d((0, 1, 0, 1))) modules.add_module('_debug_padding_%d' % i, nn.ZeroPad2d((0, 1, 0, 1)))
maxpool = nn.MaxPool2d(kernel_size=kernel_size, stride=stride, padding=int((kernel_size - 1) // 2)) maxpool = nn.MaxPool2d(kernel_size=kernel_size, stride=stride, padding=int((kernel_size - 1) // 2))
modules.add_module('maxpool_%d' % i, maxpool) modules.add_module('maxpool_%d' % i, maxpool)
elif module_def['type'] == 'upsample': elif module_def['type'] == 'upsample':
upsample = nn.Upsample(scale_factor=int(module_def['stride']), mode='nearest') modules = nn.Upsample(scale_factor=int(module_def['stride']), mode='nearest')
modules.add_module('upsample_%d' % i, upsample)
elif module_def['type'] == 'route': elif module_def['type'] == 'route': # nn.Sequential() placeholder for 'route' layer
layers = [int(x) for x in module_def['layers'].split(',')] layers = [int(x) for x in module_def['layers'].split(',')]
filters = sum([output_filters[i + 1 if i > 0 else i] for i in layers]) filters = sum([output_filters[i + 1 if i > 0 else i] for i in layers])
modules.add_module('route_%d' % i, nn.Sequential()) # Placeholder for 'route' layer
# if module_defs[i+1]['type'] == 'reorg3d': # if module_defs[i+1]['type'] == 'reorg3d':
# upsample = nn.Upsample(scale_factor=1/float(module_defs[i+1]['stride']), mode='nearest') # modules = nn.Upsample(scale_factor=1/float(module_defs[i+1]['stride']), mode='nearest') # reorg3d
# modules.add_module('reorg3d_%d' % i, upsample)
elif module_def['type'] == 'shortcut':
filters = output_filters[int(module_def['from'])]
modules.add_module('shortcut_%d' % i, nn.Sequential()) # Placeholder for 'shortcut' layer
elif module_def['type'] == 'reorg3d': elif module_def['type'] == 'shortcut': # nn.Sequential() placeholder for 'shortcut' layer
filters = output_filters[int(module_def['from'])]
elif module_def['type'] == 'reorg3d': # yolov3-spp-pan-scale
# torch.Size([16, 128, 104, 104]) # torch.Size([16, 128, 104, 104])
# torch.Size([16, 64, 208, 208]) <-- # stride 2 interpolate dimensions 2 and 3 to cat with prior layer # torch.Size([16, 64, 208, 208]) <-- # stride 2 interpolate dimensions 2 and 3 to cat with prior layer
pass pass
elif module_def['type'] == 'yolo': elif module_def['type'] == 'yolo':
yolo_index += 1 yolo_index += 1
anchor_idxs = [int(x) for x in module_def['mask'].split(',')] mask = [int(x) for x in module_def['mask'].split(',')] # anchor mask
# Extract anchors a = [float(x) for x in module_def['anchors'].split(',')] # anchors
anchors = [float(x) for x in module_def['anchors'].split(',')] a = [(a[i], a[i + 1]) for i in range(0, len(a), 2)]
anchors = [(anchors[i], anchors[i + 1]) for i in range(0, len(anchors), 2)] modules = YOLOLayer(anchors=[a[i] for i in mask], # anchor list
anchors = [anchors[i] for i in anchor_idxs] nc=int(module_def['classes']), # number of classes
nc = int(module_def['classes']) # number of classes img_size=hyperparams['height'], # 416
img_size = hyperparams['height'] yolo_index=yolo_index) # 0, 1 or 2
# Define detection layer
modules.add_module('yolo_%d' % i, YOLOLayer(anchors, nc, img_size, yolo_index))
else: else:
print('Warning: Unrecognized Layer Type: ' + module_def['type']) print('Warning: Unrecognized Layer Type: ' + module_def['type'])
@ -146,11 +141,11 @@ class YOLOLayer(nn.Module):
io[..., 0:2] = torch.sigmoid(io[..., 0:2]) + self.grid_xy # xy io[..., 0:2] = torch.sigmoid(io[..., 0:2]) + self.grid_xy # xy
io[..., 2:4] = torch.exp(io[..., 2:4]) * self.anchor_wh # wh yolo method io[..., 2:4] = torch.exp(io[..., 2:4]) * self.anchor_wh # wh yolo method
io[..., :4] *= self.stride io[..., :4] *= self.stride
# io[..., 2:4] = ((torch.sigmoid(io[..., 2:4]) * 2) ** 3) * self.anchor_wh # wh power method # io[..., 2:4] = ((torch.sigmoid(io[..., 2:4]) * 2) ** 3) * self.anchor_wh # wh power method
io[..., 4:] = torch.sigmoid(io[..., 4:]) # p_conf, p_cls io[..., 4:] = torch.sigmoid(io[..., 4:]) # p_conf, p_cls
# io[..., 5:] = F.softmax(io[..., 5:], dim=4) # p_cls # io[..., 5:] = F.softmax(io[..., 5:], dim=4) # p_cls
if self.nc == 1: if self.nc == 1:
io[..., 5] = 1 # single-class model https://github.com/ultralytics/yolov3/issues/235 io[..., 5] = 1 # single-class model https://github.com/ultralytics/yolov3/issues/235
@ -198,7 +193,7 @@ class Darknet(nn.Module):
layer_i = int(module_def['from']) layer_i = int(module_def['from'])
x = layer_outputs[-1] + layer_outputs[layer_i] x = layer_outputs[-1] + layer_outputs[layer_i]
elif mtype == 'yolo': elif mtype == 'yolo':
x = module[0](x, img_size) x = module(x, img_size)
output.append(x) output.append(x)
layer_outputs.append(x) layer_outputs.append(x)