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

updates

This commit is contained in:
Glenn Jocher 2020-02-17 15:28:11 -08:00
parent 9880dcd6cd
commit 45ce01f859
2 changed files with 11 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
models.py
View File

@ -67,9 +67,9 @@ def create_modules(module_defs, img_size, arc):
# modules = nn.Upsample(scale_factor=1/float(mdef[i+1]['stride']), mode='nearest') # reorg3d # modules = nn.Upsample(scale_factor=1/float(mdef[i+1]['stride']), mode='nearest') # reorg3d
elif mdef['type'] == 'shortcut': # nn.Sequential() placeholder for 'shortcut' layer elif mdef['type'] == 'shortcut': # nn.Sequential() placeholder for 'shortcut' layer
layer = int(mdef['from']) layers = [int(x) for x in mdef['from'].split(',')]
filters = output_filters[layer] filters = output_filters[layers[0]]
routs.extend([i + layer if layer < 0 else layer]) routs.extend([i + l if l < 0 else l for l in layers])
elif mdef['type'] == 'reorg3d': # yolov3-spp-pan-scale elif mdef['type'] == 'reorg3d': # yolov3-spp-pan-scale
# torch.Size([16, 128, 104, 104]) # torch.Size([16, 128, 104, 104])
@ -239,10 +239,10 @@ class Darknet(nn.Module):
mtype = mdef['type'] mtype = mdef['type']
if mtype in ['convolutional', 'upsample', 'maxpool']: if mtype in ['convolutional', 'upsample', 'maxpool']:
x = module(x) x = module(x)
elif mtype == 'route': elif mtype == 'route': # concat
layers = [int(x) for x in mdef['layers'].split(',')] layers = [int(x) for x in mdef['layers'].split(',')]
if verbose: if verbose:
print('route concatenating %s' % ([layer_outputs[i].shape for i in layers])) print('route/concatenate %s' % ([layer_outputs[i].shape for i in layers]))
if len(layers) == 1: if len(layers) == 1:
x = layer_outputs[layers[0]] x = layer_outputs[layers[0]]
else: else:
@ -252,10 +252,11 @@ class Darknet(nn.Module):
layer_outputs[layers[1]] = F.interpolate(layer_outputs[layers[1]], scale_factor=[0.5, 0.5]) layer_outputs[layers[1]] = F.interpolate(layer_outputs[layers[1]], scale_factor=[0.5, 0.5])
x = torch.cat([layer_outputs[i] for i in layers], 1) x = torch.cat([layer_outputs[i] for i in layers], 1)
# print(''), [print(layer_outputs[i].shape) for i in layers], print(x.shape) # print(''), [print(layer_outputs[i].shape) for i in layers], print(x.shape)
elif mtype == 'shortcut': elif mtype == 'shortcut': # sum
j = int(mdef['from']) layers = [int(x) for x in mdef['from'].split(',')]
if verbose: if verbose:
print('shortcut adding layer %g-%s to %g-%s' % (j, layer_outputs[j].shape, i - 1, x.shape)) print('shortcut/add %s' % ([layer_outputs[i].shape for i in layers]))
for j in layers:
x = x + layer_outputs[j] x = x + layer_outputs[j]
elif mtype == 'yolo': elif mtype == 'yolo':
output.append(module(x, img_size)) output.append(module(x, img_size))

2
utils/parse_config.py
View File

@ -33,7 +33,7 @@ def parse_model_cfg(path):
# Check all fields are supported # Check all fields are supported
supported = ['type', 'batch_normalize', 'filters', 'size', 'stride', 'pad', 'activation', 'layers', 'groups', supported = ['type', 'batch_normalize', 'filters', 'size', 'stride', 'pad', 'activation', 'layers', 'groups',
'from', 'mask', 'anchors', 'classes', 'num', 'jitter', 'ignore_thresh', 'truth_thresh', 'random', 'from', 'mask', 'anchors', 'classes', 'num', 'jitter', 'ignore_thresh', 'truth_thresh', 'random',
'stride_x', 'stride_y'] 'stride_x', 'stride_y', 'weights_type', 'weights_normalization']
f = [] # fields f = [] # fields
for x in mdefs[1:]: for x in mdefs[1:]: