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

updates

This commit is contained in:
Glenn Jocher 2019-06-12 13:04:58 +02:00
parent 64933f7ce0
commit 8df215a8cc
2 changed files with 20 additions and 33 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

43
train.py
View File

@ -64,7 +64,6 @@ def train(
epochs=100, # 500200 batches at bs 4, 117263 images = 68 epochs epochs=100, # 500200 batches at bs 4, 117263 images = 68 epochs
batch_size=16, batch_size=16,
accumulate=4, # effective bs = 64 = batch_size * accumulate accumulate=4, # effective bs = 64 = batch_size * accumulate
multi_scale=True,
freeze_backbone=False, freeze_backbone=False,
transfer=False # Transfer learning (train only YOLO layers) transfer=False # Transfer learning (train only YOLO layers)
): ):
@ -73,12 +72,13 @@ def train(
latest = weights + 'latest.pt' latest = weights + 'latest.pt'
best = weights + 'best.pt' best = weights + 'best.pt'
device = torch_utils.select_device() device = torch_utils.select_device()
torch.backends.cudnn.benchmark = True # unsuitable for multiscale torch.backends.cudnn.benchmark = True # possibly unsuitable for multiscale
img_size_test = img_size # image size for testing
if multi_scale: if opt.multi_scale:
min_size = round(img_size / 32 / 1.5) img_size_min = round(img_size / 32 / 1.5)
max_size = round(img_size / 32 * 1.5) img_size_max = round(img_size / 32 * 1.5)
img_size = max_size * 32 # initiate with maximum multi_scale size img_size = img_size_max * 32 # initiate with maximum multi_scale size
# opt.num_workers = 0 # bug https://github.com/ultralytics/yolov3/issues/174 # opt.num_workers = 0 # bug https://github.com/ultralytics/yolov3/issues/174
# Configure run # Configure run
@ -87,7 +87,7 @@ def train(
nc = int(data_dict['classes']) # number of classes nc = int(data_dict['classes']) # number of classes
# Initialize model # Initialize model
model = Darknet(cfg, img_size).to(device) model = Darknet(cfg).to(device)
# Optimizer # Optimizer
optimizer = optim.SGD(model.parameters(), lr=hyp['lr0'], momentum=hyp['momentum'], weight_decay=hyp['weight_decay']) optimizer = optim.SGD(model.parameters(), lr=hyp['lr0'], momentum=hyp['momentum'], weight_decay=hyp['weight_decay'])
@ -144,8 +144,7 @@ def train(
img_size, img_size,
batch_size, batch_size,
augment=True, augment=True,
rect=False, rect=False)
multi_scale=multi_scale)
# Initialize distributed training # Initialize distributed training
if torch.cuda.device_count() > 1: if torch.cuda.device_count() > 1:
@ -204,6 +203,14 @@ def train(
imgs = imgs.to(device) imgs = imgs.to(device)
targets = targets.to(device) targets = targets.to(device)
# Multi-Scale training
if opt.multi_scale:
if (i + 1 + nb * epoch) % 10 == 0: #  adjust (67% - 150%) every 10 batches
img_size = random.choice(range(img_size_min, img_size_max + 1)) * 32
print('multi_scale img_size = %g' % img_size)
scale_factor = img_size / max(imgs.shape[-2:])
imgs = F.interpolate(imgs, scale_factor=scale_factor, 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, fname='train_batch0.jpg') plot_images(imgs=imgs, targets=targets, fname='train_batch0.jpg')
@ -243,22 +250,10 @@ def train(
t = time.time() t = time.time()
print(s) print(s)
# Multi-Scale training (67% - 150%) every 10 batches
if multi_scale and (i + 1) % 10 == 0:
dataset.img_size = random.choice(range(min_size, max_size + 1)) * 32
dataloader = DataLoader(dataset,
batch_size=batch_size,
num_workers=opt.num_workers,
shuffle=True, # disable rectangular training if True
pin_memory=True,
collate_fn=dataset.collate_fn)
print('multi_scale img_size = %g' % dataset.img_size)
# Calculate mAP (always test final epoch, skip first 5 if opt.nosave) # Calculate mAP (always test final epoch, skip first 5 if opt.nosave)
if not (opt.notest or (opt.nosave and epoch < 10)) or epoch == epochs - 1: if not (opt.notest or (opt.nosave and epoch < 10)) or epoch == epochs - 1:
with torch.no_grad(): with torch.no_grad():
results, maps = test.test(cfg, data_cfg, batch_size=batch_size, img_size=img_size, model=model, results, maps = test.test(cfg, data_cfg, batch_size=batch_size, img_size=img_size_test, model=model,
conf_thres=0.1) conf_thres=0.1)
# Write epoch results # Write epoch results
@ -316,7 +311,7 @@ if __name__ == '__main__':
parser.add_argument('--accumulate', type=int, default=4, help='accumulate gradient x batches before optimizing') parser.add_argument('--accumulate', type=int, default=4, help='accumulate gradient x batches before optimizing')
parser.add_argument('--cfg', type=str, default='cfg/yolov3-spp.cfg', help='cfg file path') parser.add_argument('--cfg', type=str, default='cfg/yolov3-spp.cfg', help='cfg file path')
parser.add_argument('--data-cfg', type=str, default='data/coco_64img.data', help='coco.data file path') parser.add_argument('--data-cfg', type=str, default='data/coco_64img.data', help='coco.data file path')
parser.add_argument('--multi-scale', action='store_true', help='random image sizes per batch 320 - 608') parser.add_argument('--multi-scale', action='store_false', help='random image sizes per batch 320 - 608')
parser.add_argument('--img-size', type=int, default=416, help='inference size (pixels)') parser.add_argument('--img-size', type=int, default=416, help='inference size (pixels)')
parser.add_argument('--resume', action='store_true', help='resume training flag') parser.add_argument('--resume', action='store_true', help='resume training flag')
parser.add_argument('--transfer', action='store_true', help='transfer learning flag') parser.add_argument('--transfer', action='store_true', help='transfer learning flag')
@ -346,7 +341,6 @@ if __name__ == '__main__':
epochs=opt.epochs, epochs=opt.epochs,
batch_size=opt.batch_size, batch_size=opt.batch_size,
accumulate=opt.accumulate, accumulate=opt.accumulate,
multi_scale=opt.multi_scale,
) )
# Evolve hyperparameters (optional) # Evolve hyperparameters (optional)
@ -383,7 +377,6 @@ if __name__ == '__main__':
epochs=opt.epochs, epochs=opt.epochs,
batch_size=opt.batch_size, batch_size=opt.batch_size,
accumulate=opt.accumulate, accumulate=opt.accumulate,
multi_scale=opt.multi_scale,
) )
mutation_fitness = results[2] mutation_fitness = results[2]

10
utils/datasets.py
View File

@ -130,8 +130,7 @@ class LoadWebcam: # for inference
class LoadImagesAndLabels(Dataset): # for training/testing class LoadImagesAndLabels(Dataset): # for training/testing
def __init__(self, path, img_size=416, batch_size=16, augment=False, rect=True, image_weights=False, def __init__(self, path, img_size=416, batch_size=16, augment=False, rect=True, image_weights=False):
multi_scale=False):
with open(path, 'r') as f: with open(path, 'r') as f:
img_files = f.read().splitlines() img_files = f.read().splitlines()
self.img_files = list(filter(lambda x: len(x) > 0, img_files)) self.img_files = list(filter(lambda x: len(x) > 0, img_files))
@ -153,11 +152,6 @@ class LoadImagesAndLabels(Dataset): # for training/testing
replace('.bmp', '.txt'). replace('.bmp', '.txt').
replace('.png', '.txt') for x in self.img_files] replace('.png', '.txt') for x in self.img_files]
multi_scale = False
if multi_scale:
s = img_size / 32
self.multi_scale = ((np.linspace(0.5, 1.5, nb) * s).round().astype(np.int) * 32)
# Rectangular Training https://github.com/ultralytics/yolov3/issues/232 # Rectangular Training https://github.com/ultralytics/yolov3/issues/232
if self.rect: if self.rect:
from PIL import Image from PIL import Image
@ -256,7 +250,7 @@ class LoadImagesAndLabels(Dataset): # for training/testing
shape = self.batch_shapes[self.batch[index]] shape = self.batch_shapes[self.batch[index]]
img, ratio, padw, padh = letterbox(img, new_shape=shape, mode='rect') img, ratio, padw, padh = letterbox(img, new_shape=shape, mode='rect')
else: else:
shape = int(self.multi_scale[self.batch[index]]) if hasattr(self, 'multi_scale') else self.img_size shape = self.img_size
img, ratio, padw, padh = letterbox(img, new_shape=shape, mode='square') img, ratio, padw, padh = letterbox(img, new_shape=shape, mode='square')
# Load labels # Load labels