def run_demo(cfg, ckpt, score_threshold, images_dir, output_dir, dataset_type):
if dataset_type == "voc":
class_names = VOCDataset.class_names
elif dataset_type == 'coco':
class_names = COCODataset.class_names
else:
raise NotImplementedError('Not implemented now.')
device = torch.device(cfg.MODEL.DEVICE)
model = build_detection_model(cfg)
model = model.to(device)
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR)
checkpointer.load(ckpt, use_latest=ckpt is None)
weight_file = ckpt if ckpt else checkpointer.get_checkpoint_file()
print('Loaded weights from {}'.format(weight_file))
# dummy_input = torch.randn(1, 3, 300, 300, device='cuda')
# input_names = ["input"]
# output_names = ["output"]
# torch.onnx.export(model, dummy_input, "vgg_ssd300_voc.onnx", verbose=True, input_names=input_names, output_names=output_names)
image_paths = glob.glob(os.path.join(images_dir, '*.jpg'))
mkdir(output_dir)
cpu_device = torch.device("cpu")
transforms = build_transforms(cfg, is_train=False)
model.eval()
for i, image_path in enumerate(image_paths):
start = time.time()
image_name = os.path.basename(image_path)
image = np.array(Image.open(image_path).convert("RGB"))
height, width = image.shape[:2]
images = transforms(image)[0].unsqueeze(0)
load_time = time.time() - start
start = time.time()
result = model(images.to(device))[0]
inference_time = time.time() - start
result = result.resize((width, height)).to(cpu_device).numpy()
boxes, labels, scores = result['boxes'], result['labels'], result[
'scores']
indices = scores > score_threshold
boxes = boxes[indices]
labels = labels[indices]
scores = scores[indices]
meters = ' | '.join([
'objects {:02d}'.format(len(boxes)),
'load {:03d}ms'.format(round(load_time * 1000)),
'inference {:03d}ms'.format(round(inference_time * 1000)),
'FPS {}'.format(round(1.0 / inference_time))
])
print('({:04d}/{:04d}) {}: {}'.format(i + 1, len(image_paths),
image_name, meters))
drawn_image = draw_boxes(image, boxes, labels, scores,
class_names).astype(np.uint8)
Image.fromarray(drawn_image).save(os.path.join(output_dir, image_name))
def evaluation(cfg, ckpt, distributed):
logger = logging.getLogger("SSD.inference")
model = build_detection_model(cfg)
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR, logger=logger)
device = torch.device(cfg.MODEL.DEVICE)
model.to(device)
checkpointer.load(ckpt, use_latest=ckpt is None)
model.eval()
device = torch.device(cfg.MODEL.DEVICE)
data_loaders_val = make_data_loader(cfg,
is_train=False,
distributed=distributed)
for dataset_name, data_loader in zip(cfg.DATASETS.TEST, data_loaders_val):
res = []
for batch in data_loader:
images, targets, image_ids = batch
with torch.no_grad():
torch.cuda.synchronize(device)
start = time.time()
outputs = model(images.to(device))
torch.cuda.synchronize(device)
end = time.time()
res.append(end - start)
time_sum = 0.0
for i in res:
time_sum += i
print("FPS: %f" % (float(len(res) * cfg.TEST.BATCH_SIZE) / time_sum))
def run_demo(cfg,
ckpt,
score_threshold,
images_dir: pathlib.Path,
output_dir: pathlib.Path,
dataset_type,
num_images=None):
if dataset_type == "voc":
class_names = VOCDataset.class_names
elif dataset_type == 'coco':
class_names = COCODataset.class_names
elif dataset_type == "mnist":
class_names = MNISTDetection.class_names
elif dataset_type == "tdt4265":
class_names = TDT4265Dataset.class_names
elif dataset_type == "waymo":
class_names = WaymoDataset.class_names
else:
raise NotImplementedError('Not implemented now.')
model = SSDDetector(cfg)
model = torch_utils.to_cuda(model)
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR)
checkpointer.load(ckpt, use_latest=ckpt is None)
weight_file = ckpt if ckpt else checkpointer.get_checkpoint_file()
print('Loaded weights from {}'.format(weight_file))
image_paths = list(images_dir.glob("*.png")) + list(
images_dir.glob("*.jpg"))
output_dir.mkdir(exist_ok=True, parents=True)
transforms = build_transforms(cfg, is_train=False)
model.eval()
drawn_images = []
for i, image_path in enumerate(
tqdm.tqdm(image_paths[:num_images], desc="Predicting on images")):
image_name = image_path.stem
image = np.array(Image.open(image_path).convert("RGB"))
height, width = image.shape[:2]
images = transforms(image)[0].unsqueeze(0)
result = model(torch_utils.to_cuda(images))[0]
result = result.resize((width, height)).cpu().numpy()
boxes, labels, scores = result['boxes'], result['labels'], result[
'scores']
indices = scores > score_threshold
boxes = boxes[indices]
labels = labels[indices]
scores = scores[indices]
drawn_image = draw_boxes(image, boxes, labels, scores,
class_names).astype(np.uint8)
drawn_images.append(drawn_image)
im = Image.fromarray(drawn_image)
output_path = output_dir.joinpath(f"{image_name}.png")
im.save(output_path)
return drawn_images
def evaluation(cfg, ckpt):
logger = logging.getLogger("SSD.inference")
model = SSDDetector(cfg)
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR, logger=logger)
model = torch_utils.to_cuda(model)
checkpointer.load(ckpt, use_latest=ckpt is None)
do_evaluation(cfg, model)
def run_demo(cfg, ckpt, score_threshold, images_dir: pathlib.Path,
output_dir: pathlib.Path, dataset_type):
if dataset_type == "voc":
class_names = VOCDataset.class_names
elif dataset_type == 'coco':
class_names = COCODataset.class_names
elif dataset_type == "mnist":
class_names = MNISTDetection.class_names
else:
raise NotImplementedError('Not implemented now.')
model = SSDDetector(cfg)
model = torch_utils.to_cuda(model)
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR)
checkpointer.load(ckpt, use_latest=ckpt is None)
weight_file = ckpt if ckpt else checkpointer.get_checkpoint_file()
print('Loaded weights from {}'.format(weight_file))
image_paths = list(images_dir.glob("*.png")) + list(
images_dir.glob("*.jpg"))
output_dir.mkdir(exist_ok=True, parents=True)
transforms = build_transforms(cfg, is_train=False)
model.eval()
drawn_images = []
for i, image_path in enumerate(image_paths):
start = time.time()
image_name = image_path.name
image = np.array(Image.open(image_path).convert("RGB"))
height, width = image.shape[:2]
images = transforms(image)[0].unsqueeze(0)
load_time = time.time() - start
start = time.time()
result = model(torch_utils.to_cuda(images))[0]
inference_time = time.time() - start
result = result.resize((width, height)).cpu().numpy()
boxes, labels, scores = result['boxes'], result['labels'], result[
'scores']
indices = scores > score_threshold
boxes = boxes[indices]
labels = labels[indices]
scores = scores[indices]
meters = "|".join([
'objects {:02d}'.format(len(boxes)),
'load {:03d}ms'.format(round(load_time * 1000)),
'inference {:03d}ms'.format(round(inference_time * 1000)),
'FPS {}'.format(round(1.0 / inference_time))
])
image_name = image_path.name
drawn_image = draw_boxes(image, boxes, labels, scores,
class_names).astype(np.uint8)
drawn_images.append(drawn_image)
return drawn_images
def evaluation(cfg, ckpt, distributed):
logger = logging.getLogger("SSD.inference")
model = build_detection_model(cfg)
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR, logger=logger)
device = torch.device(cfg.MODEL.DEVICE)
model.to(device)
checkpointer.load(ckpt, use_latest=ckpt is None)
do_evaluation(cfg, model, distributed)
def creat_model(cfg, ckpt):
device = torch.device(cfg.MODEL.DEVICE)
model = build_detection_model(cfg)
model = model.to(device)
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR)
checkpointer.load(ckpt, use_latest=ckpt is None)
weight_file = ckpt if ckpt else checkpointer.get_checkpoint_file()
print('Loaded weights from {}'.format(weight_file))
return model
def evaluation(cfg, ckpt, distributed):
logger = logging.getLogger("SSD.inference")
model = build_detection_model(cfg)
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR, logger=logger)
device = torch.device(cfg.MODEL.DEVICE)
#model.load_state_dict(torch.load('outputs/vgg_ssd300_voc0712.pth'), strict=False)
model.to(device)
checkpointer.load(ckpt, use_latest=ckpt is None)
do_evaluation(cfg, model, distributed)
def evaluation(cfg, ckpt, distributed):
logger: logging.RootLogger = logging.getLogger("SSD.inference")
model = build_detection_model(cfg)
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR, logger=logger)
device = torch.device(cfg.MODEL.DEVICE)
model.to(device)
checkpointer.load(ckpt, use_latest=ckpt is None)
for scale in np.linspace(0.5, 1.0, 5):
logger.info(f"Running eval with rescale factor: {scale}")
eval_result = do_evaluation(cfg, model, distributed, rescale=scale)
def get_detections(cfg, ckpt):
model = SSDDetector(cfg)
model = torch_utils.to_cuda(model)
checkpointer = CheckPointer(cfg, model, save_dir=cfg.OUTPUT_DIR)
checkpointer.load(ckpt, use_latest=ckpt is None)
weight_file = ckpt if ckpt else checkpointer.get_checkpoint_file()
print('Loaded weights from {}'.format(weight_file))
dataset_path = DatasetCatalog.DATASETS["tdt4265_test"]["data_dir"]
dataset_path = pathlib.Path(cfg.DATASET_DIR, dataset_path)
image_dir = pathlib.Path(dataset_path, "images")
image_paths = list(image_dir.glob("*.jpg"))
transforms = build_transforms(cfg, is_train=False)
model.eval()
detections = []
labels = read_labels(
image_dir.parent.parent.joinpath("train", "labels.json"))
check_all_images_exists(labels, image_paths)
# Filter labels on if they are test and only take the 7th frame
labels = [label for label in labels if label["is_test"]]
labels = [label for label in labels if label["image_id"] % 7 == 0]
for i, label in enumerate(tqdm.tqdm(labels, desc="Inference on images")):
image_id = label["image_id"]
image_path = image_dir.joinpath(f"{image_id}.jpg")
image_detections = {"image_id": int(image_id), "bounding_boxes": []}
image = np.array(Image.open(image_path).convert("RGB"))
height, width = image.shape[:2]
images = transforms(image)[0].unsqueeze(0)
result = model(torch_utils.to_cuda(images))[0]
result = result.resize((width, height)).cpu().numpy()
boxes, labels, scores = result['boxes'], result['labels'], result[
'scores']
for idx in range(len(boxes)):
box = boxes[idx]
label_id = labels[idx]
label = TDT4265Dataset.class_names[label_id]
assert label != "__background__"
score = float(scores[idx])
assert box.shape == (4, )
json_box = {
"xmin": float(box[0]),
"ymin": float(box[1]),
"xmax": float(box[2]),
"ymax": float(box[3]),
"label": str(label),
"label_id": int(label_id),
"confidence": float(score)
}
image_detections["bounding_boxes"].append(json_box)
detections.append(image_detections)
return detections
def evaluation(cfg, ckpt, distributed, model_path=None):
logger = logging.getLogger("SSD.inference")
model = build_detection_model(cfg)
logger.info("Model :\n{}".format(model)) #如果用print,多gpu会打印两便
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR, logger=logger)
device = torch.device(cfg.MODEL.DEVICE)
model.to(device)
if model_path is None:
checkpointer.load(ckpt, use_latest=ckpt is None)
else:
model.load_state_dict(torch.load(model_path))
if cfg.TEST.BN_FUSE is True:
print('BN_FUSE.')
model.backbone.bn_fuse()
model.to(device)
do_evaluation(cfg, model, distributed)
def start_train(cfg):
logger = logging.getLogger('SSD.trainer')
model = SSDDetector(cfg)
model = torch_utils.to_cuda(model)
lr = cfg.SOLVER.LR
optimizer = make_optimizer(cfg, model, lr)
milestones = [step for step in cfg.SOLVER.LR_STEPS]
scheduler = make_lr_scheduler(cfg, optimizer, milestones)
arguments = {"iteration": 0}
save_to_disk = True
checkpointer = CheckPointer(cfg, model, optimizer, scheduler,
cfg.OUTPUT_DIR, save_to_disk, logger)
extra_checkpoint_data = checkpointer.load()
arguments.update(extra_checkpoint_data)
max_iter = cfg.SOLVER.MAX_ITER
train_loader = make_data_loader(cfg,
is_train=True,
max_iter=max_iter,
start_iter=arguments['iteration'])
model = do_train(cfg, model, train_loader, optimizer, scheduler,
checkpointer, arguments)
return model
def train(cfg, args):
# 工厂模式，加载日志文件设置，这里暂时不同管
logger = logging.getLogger('SSD.trainer')
# 建立目标检测模型
model = build_detection_model(cfg)
# 设置Device并且把模型部署到设备上
device = torch.device(cfg.MODEL.DEVICE)
model.to(device)
if args.distributed:
model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.local_rank], output_device=args.local_rank)
# 设置学习率、优化器还有学习率变化步长，可以理解为模拟退火这种，前面的步长比较大，后面的步长比较小
lr = cfg.SOLVER.LR * args.num_gpus # scale by num gpus
optimizer = make_optimizer(cfg, model, lr)
milestones = [step // args.num_gpus for step in cfg.SOLVER.LR_STEPS]
scheduler = make_lr_scheduler(cfg, optimizer, milestones)
arguments = {"iteration": 0}
save_to_disk = dist_util.get_rank() == 0
# **** 这里应该是从断点开始对模型进行训练 ****
checkpointer = CheckPointer(model, optimizer, scheduler, cfg.OUTPUT_DIR, save_to_disk, logger)
extra_checkpoint_data = checkpointer.load()
arguments.update(extra_checkpoint_data)
# Important 通过torch的形式去加载数据集
# 关键在于如何加载数据集，模型的构建过程可以简单地看成是黑盒
max_iter = cfg.SOLVER.MAX_ITER // args.num_gpus
train_loader = make_data_loader(cfg, is_train=True, distributed=args.distributed, max_iter=max_iter, start_iter=arguments['iteration'])
# 正式开始训练, 暂时先不训练？
# 不对，不训练也得加载数据集**** 暂时不训练就完事了 *** 直接看数据加载过程
# model = do_train(cfg, model, train_loader, optimizer, scheduler, checkpointer, device, arguments, args)
return model
def main(video, config):
class_name = ('__background__', 'lubang', 'retak aligator',
'retak melintang', 'retak memanjang')
cfg.merge_from_file(config)
cfg.freeze()
ckpt = None
device = torch.device(cfg.MODEL.DEVICE)
model = build_detection_model(cfg)
model.to(device)
checkpoint = CheckPointer(model, save_dir=cfg.OUTPUT_DIR)
checkpoint.load(ckpt, use_latest=ckpt is None)
weight_file = ckpt if ckpt else checkpoint.get_checkpoint_file()
print(f'Loading weight from {weight_file}')
def start_train(cfg):
logger = logging.getLogger('SSD.trainer')
model = SSDDetector(cfg)
model = torch_utils.to_cuda(model)
optimizer = torch.optim.SGD(
filter(lambda p: p.requires_grad, model.parameters()),
lr=cfg.SOLVER.LR,
momentum=cfg.SOLVER.MOMENTUM,
weight_decay=cfg.SOLVER.WEIGHT_DECAY,
nesterov=True,
)
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(
optimizer, T_max=int(cfg.SOLVER.MAX_ITER / 1000), eta_min=0)
arguments = {"iteration": 0}
save_to_disk = True
checkpointer = CheckPointer(
model,
optimizer,
cfg.OUTPUT_DIR,
save_to_disk,
logger,
)
extra_checkpoint_data = checkpointer.load()
arguments.update(extra_checkpoint_data)
max_iter = cfg.SOLVER.MAX_ITER
train_loader = make_data_loader(cfg,
is_train=True,
max_iter=max_iter,
start_iter=arguments['iteration'])
model = do_train(cfg, model, train_loader, optimizer, checkpointer,
arguments, scheduler)
return model
def train(cfg, args):
logger = logging.getLogger('SSD.trainer')
model = build_detection_model(cfg)
device = torch.device(cfg.MODEL.DEVICE)
model.to(device)
if args.distributed:
model = torch.nn.parallel.DistributedDataParallel(
model, device_ids=[args.local_rank], output_device=args.local_rank)
lr = cfg.SOLVER.LR * args.num_gpus # scale by num gpus
optimizer = make_optimizer(cfg, model, lr)
milestones = [step // args.num_gpus for step in cfg.SOLVER.LR_STEPS]
scheduler = make_lr_scheduler(cfg, optimizer, milestones)
arguments = {"iteration": 0}
save_to_disk = dist_util.get_rank() == 0
checkpointer = CheckPointer(model, optimizer, scheduler, cfg.OUTPUT_DIR,
save_to_disk, logger)
extra_checkpoint_data = checkpointer.load()
arguments.update(extra_checkpoint_data)
max_iter = cfg.SOLVER.MAX_ITER // args.num_gpus
train_loader = make_data_loader(cfg,
is_train=True,
distributed=args.distributed,
max_iter=max_iter,
start_iter=arguments['iteration'])
model = do_train(cfg, model, train_loader, optimizer, scheduler,
checkpointer, device, arguments, args)
return model
def start_train(cfg):
logger = logging.getLogger('SSD.trainer')
model = SSDDetector(cfg)
model = torch_utils.to_cuda(model)
optimizer = torch.optim.SGD(
model.parameters(),
lr=cfg.SOLVER.LR,
momentum=cfg.SOLVER.MOMENTUM,
weight_decay=cfg.SOLVER.WEIGHT_DECAY
)
arguments = {"iteration": 0}
save_to_disk = True
checkpointer = CheckPointer(
model, optimizer, cfg.OUTPUT_DIR, save_to_disk, logger,
)
extra_checkpoint_data = checkpointer.load()
arguments.update(extra_checkpoint_data)
max_iter = cfg.SOLVER.MAX_ITER
train_loader = make_data_loader(cfg, is_train=True, max_iter=max_iter, start_iter=arguments['iteration'])
model = do_train(
cfg, model, train_loader, optimizer,
checkpointer, arguments)
return model
def train(cfg, args):
logger = logging.getLogger('SSD.trainer')
model = build_detection_model(cfg)
device = torch.device(cfg.MODEL.DEVICE)
model.to(device)
if args.distributed:
model = torch.nn.parallel.DistributedDataParallel(
model, device_ids=[args.local_rank], output_device=args.local_rank)
lr = cfg.SOLVER.LR * args.num_gpus # scale by num gpus
optimizer = make_optimizer(cfg, model, lr)
milestones = [step // args.num_gpus for step in cfg.SOLVER.LR_STEPS]
scheduler = make_lr_scheduler(cfg, optimizer, milestones)
arguments = {"iteration": 0}
save_to_disk = dist_util.get_rank() == 0
checkpointer = CheckPointer(model, optimizer, scheduler, cfg.OUTPUT_DIR,
save_to_disk, logger)
extra_checkpoint_data = checkpointer.load()
arguments.update(extra_checkpoint_data)
max_iter = cfg.SOLVER.MAX_ITER // args.num_gpus
train_loader = make_data_loader(cfg,
is_train=True,
distributed=args.distributed,
max_iter=max_iter,
start_iter=arguments['iteration'])
# macs, params = profile(model, inputs=(input, ))
#
# macs, params = clever_format([flops, params], "%.3f")
# net = model.to()
# with torch.cuda.device(0):
# net = model.to(device)
# macs, params = get_model_complexity_info(net, (3, 512, 512), as_strings=True,
# print_per_layer_stat=True, verbose=True)
# print('{:<30} {:<8}'.format('Computational complexity: ', macs))
# print('{:<30} {:<8}'.format('Number of parameters: ', params))
n_params = sum(p.numel() for name, p in model.named_parameters()
if p.requires_grad)
print(n_params)
#
# model = net
# inputs = torch.randn(1, 3, 300, 300) #8618 305
# inputs = torch.randn(1, 3, 300, 300)
# macs = profile_macs(model, inputs)
# print(macs)
model = do_train(cfg, model, train_loader, optimizer, scheduler,
checkpointer, device, arguments, args)
return model
def start_train(cfg, visualize_example=False):
logger = logging.getLogger('SSD.trainer')
model = SSDDetector(cfg)
print(model)
model = torch_utils.to_cuda(model)
optimizer = torch.optim.SGD(model.parameters(),
lr=cfg.SOLVER.LR,
momentum=cfg.SOLVER.MOMENTUM,
weight_decay=cfg.SOLVER.WEIGHT_DECAY)
"""
optimizer = torch.optim.Adam(
model.parameters(),
lr=cfg.SOLVER.LR,
weight_decay=cfg.SOLVER.WEIGHT_DECAY
)
"""
"""
lr_scheduler = torch.optim.lr_scheduler.CyclicLR(
optimizer= optimizer,
base_lr= cfg.SOLVER.LR /10,
max_lr=0.05,
step_size_up=8000,
mode='triangular2'
)
"""
arguments = {"iteration": 0}
save_to_disk = True
checkpointer = CheckPointer(
model,
optimizer,
cfg.OUTPUT_DIR,
save_to_disk,
logger,
)
extra_checkpoint_data = checkpointer.load()
arguments.update(extra_checkpoint_data)
max_iter = cfg.SOLVER.MAX_ITER
train_loader = make_data_loader(cfg,
is_train=True,
max_iter=max_iter,
start_iter=arguments['iteration'])
model = do_train(cfg,
model,
train_loader,
optimizer,
checkpointer,
arguments,
visualize_example,
lr_scheduler=None)
return model
def get_detections(cfg, ckpt):
model = SSDDetector(cfg)
model = torch_utils.to_cuda(model)
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR)
checkpointer.load(ckpt, use_latest=ckpt is None)
weight_file = ckpt if ckpt else checkpointer.get_checkpoint_file()
print('Loaded weights from {}'.format(weight_file))
dataset_path = DatasetCatalog.DATASETS["tdt4265_test"]["data_dir"]
dataset_path = pathlib.Path(cfg.DATASET_DIR, dataset_path)
image_dir = pathlib.Path(dataset_path)
image_paths = list(image_dir.glob("*.jpg"))
transforms = build_transforms(cfg, is_train=False)
model.eval()
detections = []
for image_path in tqdm.tqdm(image_paths, desc="Inference on images"):
image = np.array(Image.open(image_path).convert("RGB"))
height, width = image.shape[:2]
images = transforms(image)[0].unsqueeze(0)
result = model(torch_utils.to_cuda(images))[0]
result = result.resize((width, height)).cpu().numpy()
boxes, labels, scores = result['boxes'], result['labels'], result[
'scores']
for idx in range(len(boxes)):
box = boxes[idx]
label_id = labels[idx]
label = TDT4265Dataset.class_names[label_id]
assert label != "__background__"
score = float(scores[idx])
assert box.shape == (4, )
xmin, ymin, xmax, ymax = box
width = xmax - xmin
height = ymax - ymin
detections.append({
"image_id": image_path.stem,
"category_id": LABEL_MAP[label],
"score": score,
"bbox": [xmin, ymin, width, height]
})
return detections
def evaluation(cfg, ckpt, N_images: int):
model = SSDDetector(cfg)
logger = logging.getLogger("SSD.inference")
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR, logger=logger)
model = torch_utils.to_cuda(model)
checkpointer.load(ckpt, use_latest=ckpt is None)
model.eval()
data_loaders_val = make_data_loader(cfg, is_train=False)
for data_loader in data_loaders_val:
batch = next(iter(data_loader))
images, targets, image_ids = batch
images = torch_utils.to_cuda(images)
imshape = list(images.shape[2:])
# warmup
print("Checking runtime for image shape:", imshape)
for i in range(10):
model(images)
start_time = time.time()
for i in range(N_images):
outputs = model(images)
total_time = time.time() - start_time
print("Runtime for image shape:", imshape)
print("Total runtime:", total_time)
print("FPS:", N_images / total_time)
def start_train(cfg):
logger = logging.getLogger('SSD.trainer')
model = SSDDetector(cfg)
model = torch_utils.to_cuda(model)
if cfg.SOLVER.TYPE == "adam":
optimizer = torch.optim.Adam(
model.parameters(),
lr=cfg.SOLVER.LR,
weight_decay=cfg.SOLVER.WEIGHT_DECAY,
)
elif cfg.SOLVER.TYPE == "sgd":
optimizer = torch.optim.SGD(model.parameters(),
lr=cfg.SOLVER.LR,
weight_decay=cfg.SOLVER.WEIGHT_DECAY,
momentum=cfg.SOLVER.MOMENTUM)
else:
# Default to Adam if incorrect solver
print("WARNING: Incorrect solver type, defaulting to Adam")
optimizer = torch.optim.Adam(
model.parameters(),
lr=cfg.SOLVER.LR,
weight_decay=cfg.SOLVER.WEIGHT_DECAY,
)
scheduler = LinearMultiStepWarmUp(cfg, optimizer)
arguments = {"iteration": 0}
save_to_disk = True
checkpointer = CheckPointer(
model,
optimizer,
cfg.OUTPUT_DIR,
save_to_disk,
logger,
)
extra_checkpoint_data = checkpointer.load()
arguments.update(extra_checkpoint_data)
max_iter = cfg.SOLVER.MAX_ITER
train_loader = make_data_loader(cfg,
is_train=True,
max_iter=max_iter,
start_iter=arguments['iteration'])
model = do_train(cfg, model, train_loader, optimizer, checkpointer,
arguments, scheduler)
return model
def train(cfg, args):
logger = logging.getLogger('SSD.trainer')
model = build_detection_model(cfg)
device = torch.device(cfg.MODEL.DEVICE)
model.to(device)
if args.distributed:
model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model)
model = torch.nn.parallel.DistributedDataParallel(
model, device_ids=[args.local_rank], output_device=args.local_rank)
lr = cfg.SOLVER.LR * args.num_gpus # scale by num gpus
optimizer = make_optimizer(cfg, model, lr)
milestones = [step // args.num_gpus for step in cfg.SOLVER.LR_STEPS]
scheduler = make_lr_scheduler(cfg, optimizer, milestones)
arguments = {"iteration": 0}
save_to_disk = dist_util.get_rank() == 0
checkpointer = CheckPointer(model, optimizer, scheduler, cfg.OUTPUT_DIR,
save_to_disk, logger)
extra_checkpoint_data = checkpointer.load(args.ckpt)
arguments.update(extra_checkpoint_data)
max_iter = cfg.SOLVER.MAX_ITER // args.num_gpus
train_loader = make_data_loader(cfg,
is_train=True,
distributed=args.distributed,
max_iter=max_iter,
start_iter=arguments['iteration'])
logging.info('==>Start statistic')
do_run(cfg, model, distributed=args.distributed)
logging.info('==>End statistic')
for ops in model.modules():
if isinstance(ops, torch.nn.ReLU):
ops.collectStats = False
# ops.c.data = ops.running_mean + (ops.running_b * laplace[args.actBitwidth])
ops.c.data = ops.running_mean + (3 * ops.running_std)
ops.quant = True
torch.cuda.empty_cache()
model = do_train(cfg, model, train_loader, optimizer, scheduler,
checkpointer, device, arguments, args)
return model
def start_train(cfg):
logger = logging.getLogger('SSD.trainer')
model = SSDDetector(cfg)
model = torch_utils.to_cuda(model)
# SGD
# optimizer = torch.optim.SGD(
# model.parameters(),
# lr=cfg.SOLVER.LR,
# momentum=cfg.SOLVER.MOMENTUM,
# weight_decay=cfg.SOLVER.WEIGHT_DECAY
# )
# Adam
optimizer = torch.optim.Adam(model.parameters(),
lr=cfg.SOLVER.LR,
weight_decay=cfg.SOLVER.WEIGHT_DECAY)
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer=optimizer,
milestones=[6000, 10000],
gamma=cfg.SOLVER.GAMMA)
arguments = {"iteration": 0}
save_to_disk = True
checkpointer = CheckPointer(
model,
optimizer,
cfg.OUTPUT_DIR,
save_to_disk,
logger,
)
extra_checkpoint_data = checkpointer.load()
arguments.update(extra_checkpoint_data)
max_iter = cfg.SOLVER.MAX_ITER
train_loader = make_data_loader(cfg,
is_train=True,
max_iter=max_iter,
start_iter=arguments['iteration'])
model = do_train(cfg, model, train_loader, optimizer, checkpointer,
arguments, scheduler)
return model
def train(cfg: CfgNode,
args: Namespace,
output_dir: Path,
model_manager: Dict[str, Any],
freeze_non_sigma: bool = False):
logger = logging.getLogger('SSD.trainer')
model = build_detection_model(cfg)
device = torch.device(cfg.MODEL.DEVICE)
model.to(device)
if args.distributed:
model = torch.nn.parallel.DistributedDataParallel(
model, device_ids=[args.local_rank], output_device=args.local_rank)
lr = cfg.SOLVER.LR * args.num_gpus # scale by num gpus
optimizer = make_optimizer(cfg, model, lr)
milestones = [step // args.num_gpus for step in cfg.SOLVER.LR_STEPS]
scheduler = make_lr_scheduler(cfg, optimizer, milestones)
arguments = {"iteration": 0}
save_to_disk = dist_util.get_rank() == 0
checkpointer = CheckPointer(model, optimizer, scheduler, cfg.OUTPUT_DIR,
save_to_disk, logger)
resume_from = checkpointer.get_best_from_experiment_dir(cfg)
extra_checkpoint_data = checkpointer.load(f=resume_from)
arguments.update(extra_checkpoint_data)
max_iter = cfg.SOLVER.MAX_ITER // args.num_gpus
train_loader = make_data_loader(cfg,
is_train=True,
distributed=args.distributed,
max_iter=max_iter,
start_iter=arguments['iteration'])
# Weight freezing test:
# print_model(model)
# freeze_weights(model)
print_model(model)
model = do_train(cfg, model, train_loader, optimizer, scheduler,
checkpointer, device, arguments, args, output_dir,
model_manager)
return model
def train(cfg, args):
logger = logging.getLogger('SSD.trainer')
model = build_detection_model(cfg) # 建立模型
device = torch.device(cfg.MODEL.DEVICE) # 看cfg怎么组织的，把文件和args剥离开
model.to(device)
if args.distributed:
model = torch.nn.parallel.DistributedDataParallel(
model, device_ids=[args.local_rank], output_device=args.local_rank)
# model = nn.DataParallel(model)
lr = cfg.SOLVER.LR * args.num_gpus # scale by num gpus
optimizer = make_optimizer(cfg, model, lr) # 建立优化器
milestones = [step // args.num_gpus for step in cfg.SOLVER.LR_STEPS]
scheduler = make_lr_scheduler(cfg, optimizer, milestones)
arguments = {"iteration": 0}
save_to_disk = dist_util.get_rank() == 0
checkpointer = CheckPointer(model,
optimizer,
scheduler,
save_dir=cfg.OUTPUT_DIR,
save_to_disk=save_to_disk,
logger=logger)
# 建立模型存储载入类,给save_dir赋值表示
extra_checkpoint_data = checkpointer.load(f='', use_latest=False) # 载入模型
arguments.update(extra_checkpoint_data)
max_iter = cfg.SOLVER.MAX_ITER // args.num_gpus
train_loader = make_data_loader(cfg,
is_train=True,
distributed=args.distributed,
max_iter=max_iter,
start_iter=arguments['iteration']) # 建立数据库
print("dataloader: ", train_loader.batch_size)
# exit(1232)
model = do_train(cfg, model, train_loader, optimizer, scheduler,
checkpointer, device, arguments, args) # 训练
return model
def run_demo(cfg, ckpt, score_threshold, images_dir, output_dir, onnx_dir, dataset_type):
if dataset_type == "voc":
class_names = VOCDataset.class_names
elif dataset_type == 'coco':
class_names = COCODataset.class_names
else:
raise NotImplementedError('Not implemented now.')
device = torch.device(cfg.MODEL.DEVICE)
device = "cpu" if not torch.cuda.is_available() else device
model = build_detection_model(cfg)
model = model.to(device)
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR)
checkpointer.load(ckpt, use_latest=ckpt is None)
weight_file = ckpt if ckpt else checkpointer.get_checkpoint_file()
print('Loaded weights from {}'.format(weight_file))
image_paths = glob.glob(os.path.join(images_dir, '*.jpg'))
mkdir(output_dir)
cpu_device = torch.device("cpu")
transforms = build_transforms(cfg, is_train=False)
model.eval()
# get model ready for onnx export
mkdir(onnx_dir)
model_onnx = build_detection_model(cfg)
model_onnx = model_onnx.to(device)
checkpointer_onnx = CheckPointer(model_onnx, save_dir=cfg.OUTPUT_DIR)
checkpointer_onnx.load(ckpt, use_latest=ckpt is None)
# replace the SSD box head postprocessor with the onnx version for exporting
model_onnx.box_head.post_processor = PostProcessorOnnx(cfg)
model_onnx.eval()
# export with ONNX
# onnx modle takes the name of the pth ckpt file
model_onnx_name = os.path.basename(ckpt).split('.')[0] + ".onnx"
model_onnx_path = os.path.join(onnx_dir, model_onnx_name)
if not os.path.exists(model_onnx_path):
print(f'Model exported as onnx to {model_onnx_path}')
dummy_input = torch.zeros(
[1, 3, cfg.INPUT.IMAGE_SIZE, cfg.INPUT.IMAGE_SIZE]).to(device)
torch.onnx.export(model_onnx,
dummy_input,
model_onnx_path,
export_params=True,
do_constant_folding=True,
opset_version=11,
input_names=['input'],
output_names=['boxes', 'scores', 'labels'],
dynamic_axes={
'input': {0: 'batch_size', 2: "height", 3: "width"}},
verbose=False)
# load exported onnx model for inference test
print(
f'Loading exported onnx model from {model_onnx_path} for inference comparison test')
onnx_runtime_sess = onnxruntime.InferenceSession(model_onnx_path)
for i, image_path in enumerate(image_paths):
start = time.time()
image_name = os.path.basename(image_path)
image = np.array(Image.open(image_path).convert("RGB"))
height, width = image.shape[:2]
images = transforms(image)[0].unsqueeze(0)
load_time = time.time() - start
start = time.time()
result = model(images.to(device))[0]
inference_time = time.time() - start
result = result.resize((width, height)).to(cpu_device).numpy()
boxes, labels, scores = result['boxes'], result['labels'], result['scores']
indices = scores > score_threshold
boxes, labels, scores = boxes[indices], labels[indices], scores[indices]
meters = ' | '.join(
[
'objects {:02d}'.format(len(boxes)),
'load {:03d}ms'.format(round(load_time * 1000)),
'inference {:03d}ms'.format(round(inference_time * 1000)),
'FPS {}'.format(round(1.0 / inference_time))
]
)
print('Pytorch: ({:04d}/{:04d}) {}: {}'.format(i + 1,
len(image_paths), image_name, meters))
drawn_image = draw_boxes(image, boxes, labels,
scores, class_names).astype(np.uint8)
Image.fromarray(drawn_image).save(
os.path.join(output_dir, "pytorch_" + image_name))
"""
Compute ONNX Runtime output prediction
"""
start = time.time()
ort_inputs = {onnx_runtime_sess.get_inputs()[0].name: np.array(images)}
boxes, scores, labels = onnx_runtime_sess.run(None, ort_inputs)
inference_time = time.time() - start
indices = scores > score_threshold
boxes, labels, scores = boxes[indices], labels[indices], scores[indices]
# resize bounding boxes to size of the original image
boxes[:, 0::2] *= (width)
boxes[:, 1::2] *= (height)
meters = ' | '.join(
[
'objects {:02d}'.format(len(boxes)),
'load {:03d}ms'.format(round(load_time * 1000)),
'inference {:03d}ms'.format(round(inference_time * 1000)),
'FPS {}'.format(round(1.0 / inference_time))
]
)
print('Onnx: ({:04d}/{:04d}) {}: {}'.format(i + 1,
len(image_paths),
image_name, meters))
drawn_image = draw_boxes(image, boxes, labels,
scores, class_names).astype(np.uint8)
Image.fromarray(drawn_image).save(
os.path.join(output_dir, "onnx_" + image_name))
def run_demo(cfg, ckpt, score_threshold, images_dir, output_dir):
class_names = VOCDataset.class_names
device = torch.device(cfg.MODEL.DEVICE)
model = build_detection_model(cfg)
model = model.to(device)
checkpointer = CheckPointer(model, save_dir=cfg.OUTPUT_DIR)
checkpointer.load(ckpt, use_latest=ckpt is None)
weight_file = ckpt if ckpt else checkpointer.get_checkpoint_file()
print('Loaded weights from {}'.format(weight_file))
image_paths = glob.glob(os.path.join(images_dir, '*.bmp'))
mkdir(output_dir)
cpu_device = torch.device("cpu")
transforms = build_transforms(cfg, is_train=False)
model.eval()
for i, image_path in enumerate(image_paths):
start = time.time()
image_name = os.path.basename(image_path)
image = np.array(Image.open(image_path).convert("RGB"))
height, width = image.shape[:2]
images = transforms(image)[0].unsqueeze(0)
load_time = time.time() - start
start = time.time()
result = model(images.to(device))[0]
inference_time = time.time() - start
result = result.resize((width, height)).to(cpu_device).numpy()
boxes, labels, scores = result['boxes'], result['labels'], result[
'scores']
indices = scores > score_threshold
boxes = boxes[indices]
labels = labels[indices]
meters = ' | '.join([
'objects {:02d}'.format(len(boxes)),
'load {:03d}ms'.format(round(load_time * 1000)),
'inference {:03d}ms'.format(round(inference_time * 1000)),
'FPS {}'.format(round(1.0 / inference_time))
])
print('({:04d}/{:04d}) {}: {}'.format(i + 1, len(image_paths),
image_name, meters))
text = ['__background__']
resDic = {}
for j in range(len(boxes)):
xmin = int(boxes[j, 0])
ymin = int(boxes[j, 1])
xmax = int(boxes[j, 2])
ymax = int(boxes[j, 3])
if labels[j] == 1:
xmin += 140
xmax -= 130
elif labels[j] == 2:
xmin += 130
elif labels[j] == 4:
xmin += 40
hight = ymax - ymin
width = xmax - xmin
cropImg = image[ymin:ymin + hight, xmin:xmin + width]
cropImg = local_threshold(cropImg)
boxes[j, 0] = xmin
boxes[j, 1] = ymin
boxes[j, 2] = xmax
boxes[j, 3] = ymax
text_tmp = crnnOcr(Image.fromarray(cropImg))
if labels[j] == 2:
text_tmp = re.sub('[^\x00-\xff]', '/', text_tmp)
text.append(text_tmp)
resDic[class_names[labels[j]]] = text_tmp
result = json.dumps(resDic, ensure_ascii=False)
print(result)
def get_model(self, cfg, weightfile):
model = build_detection_model(cfg)
model = model.to(self.device)
checkpointer = CheckPointer(model)
checkpointer.load(weightfile, use_latest=weightfile is None)
return model
result_file = './results/feature_maps_frame75.jpg'
class_name = {
'__background__', 'lubang', 'retak aligator', 'retak melintang',
'retak memanjang'
}
cfg.merge_from_file(config)
cfg.freeze()
ckpt = None
device = torch.device('cpu')
model = build_detection_model(cfg)
model.to(device)
checkpoint = CheckPointer(model, save_dir=cfg.OUTPUT_DIR)
checkpoint.load(ckpt, use_latest=ckpt is None)
weight_file = ckpt if ckpt else checkpoint.get_checkpoint_file()
transforms = build_transforms(cfg, is_train=False)
model.eval()
conv_layers = []
model_children = list(model.children())
print(len(model_children))
print(type(model_children[0]))
print(type(model_children[1]))
counter = 0
for i in range(len(model_children)):
if type(model_children[i]) == VGG: