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

updates

This commit is contained in:
Glenn Jocher 2019-08-08 19:49:15 +02:00
parent 37c3e762e1
commit c49fe688b7
2 changed files with 9 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
models.py
View File

@ -31,7 +31,7 @@ def create_modules(module_defs):
padding=pad, padding=pad,
bias=not bn)) bias=not bn))
if bn: if bn:
modules.add_module('BatchNorm2d', nn.BatchNorm2d(filters)) modules.add_module('BatchNorm2d', nn.BatchNorm2d(filters, momentum=0.1))
if mdef['activation'] == 'leaky': if mdef['activation'] == 'leaky':
# modules.add_module('activation', nn.PReLU(num_parameters=filters, init=0.1)) # modules.add_module('activation', nn.PReLU(num_parameters=filters, init=0.1))
modules.add_module('activation', nn.LeakyReLU(0.1, inplace=True)) modules.add_module('activation', nn.LeakyReLU(0.1, inplace=True))

13
train.py
View File

@ -186,7 +186,6 @@ def train(cfg,
nb = len(dataloader) nb = len(dataloader)
maps = np.zeros(nc) # mAP per class maps = np.zeros(nc) # mAP per class
results = (0, 0, 0, 0, 0, 0, 0) # P, R, mAP, F1, test_loss results = (0, 0, 0, 0, 0, 0, 0) # P, R, mAP, F1, test_loss
# n_burnin = min(round(nb / 5 + 1), 1000) # burn-in batches
t0 = time.time() t0 = time.time()
for epoch in range(start_epoch, epochs): for epoch in range(start_epoch, epochs):
model.train() model.train()
@ -215,22 +214,26 @@ def train(cfg,
imgs = imgs.to(device) imgs = imgs.to(device)
targets = targets.to(device) targets = targets.to(device)
# Multi-Scale training TODO: short-side to 32-multiple https://github.com/ultralytics/yolov3/issues/358 # Multi-Scale training
if multi_scale: if multi_scale:
if (i + nb * epoch) / accumulate % 10 == 0: #  adjust (67% - 150%) every 10 batches if (i + nb * epoch) / accumulate % 10 == 0: #  adjust (67% - 150%) every 10 batches
img_size = random.randrange(img_sz_min, img_sz_max + 1) * 32 img_size = random.randrange(img_sz_min, img_sz_max + 1) * 32
sf = img_size / max(imgs.shape[2:]) # scale factor sf = img_size / max(imgs.shape[2:]) # scale factor
if sf != 1: if sf != 1:
ns = [math.ceil(x * sf / 32.) * 32 for x in imgs.shape[2:]] # new shape ns = [math.ceil(x * sf / 32.) * 32 for x in imgs.shape[2:]] # new shape (stretched to 32-multiple)
imgs = F.interpolate(imgs, size=ns, mode='bilinear', align_corners=False) imgs = F.interpolate(imgs, size=ns, mode='bilinear', align_corners=False)
# Plot images with bounding boxes # Plot images with bounding boxes
if epoch == 0 and i == 0: if epoch == 0 and i == 0:
plot_images(imgs=imgs, targets=targets, paths=paths, fname='train_batch%g.jpg' % i) plot_images(imgs=imgs, targets=targets, paths=paths, fname='train_batch%g.jpg' % i)
# SGD burn-in # Hyperparameter burn-in
# n_burnin = min(round(nb / 5 + 1), 1000) # burn-in batches
# if epoch == 0 and i <= n_burnin: # if epoch == 0 and i <= n_burnin:
# g = (i / n_burnin) ** 4 # gain # for m in model.named_modules():
# if m[0].endswith('BatchNorm2d'):
# m[1].momentum = 1 - i / n_burnin * 0.99 # BatchNorm2d momentum falls from 1 - 0.01
# g = (i / n_burnin) ** 4 # gain rises from 0 - 1
# for x in optimizer.param_groups: # for x in optimizer.param_groups:
# x['lr'] = hyp['lr0'] * g # x['lr'] = hyp['lr0'] * g
# x['weight_decay'] = hyp['weight_decay'] * g # x['weight_decay'] = hyp['weight_decay'] * g