我正在尝试根据PyTorch的“微调目标检测”官方教程实现一个检测模型。使用少量数据（10张图像）似乎已经成功了。但是，我将整个数据集上传到Drive并检查了索引-数据-标签对应关系。在我的设置中没有不匹配的项目，我已经解决了该部分中的所有错误。（我从GDrive标签中删除了额外的项目）
class SomeDataset(torch.utils.data.Dataset):
def __init__(self, root_path, transforms):
self.root_path = root_path
self.transforms = transforms
# load all image files, sorting them to
# ensure that they are aligned
self.imgs = list(sorted(os.listdir(os.path.join(root_path, "images"))))
self.labels = list(sorted(os.listdir(os.path.join(root_path, "labels"))))
def __getitem__(self, idx):
# load images ad masks
img_path = os.path.join(self.root_path, "images", self.imgs[idx])
label_path = os.path.join(self.root_path, "labels", self.labels[idx])
img = Image.open(img_path).convert("RGB")
# get labels and boxes
label_data = np.loadtxt(label_path, dtype=str, delimiter=' ');
print(f"{len(label_data)} is the length of label data")
num_objs = label_data.shape[0];
if num_objs != 0:
print(f"number of objects {num_objs}")
# label values should start from 1
for i,label_name in enumerate(classnames):
label_data[np.where(label_name==label_data)] = i;
label_data = label_data.astype(np.float);
print(f"label data {label_data}")
xs = label_data[:,0:8:2];
ys = label_data[:,1:8:2];
x_min = np.min(xs, axis=1)[...,np.newaxis];
x_max = np.max(xs, axis=1)[...,np.newaxis];
y_min = np.min(ys, axis=1)[...,np.newaxis];
y_max = np.max(ys, axis=1)[...,np.newaxis];
boxes = np.hstack((x_min,y_min,x_max,y_max));
labels = label_data[:,8];
else:
# if there is no label add background whose label is 0
boxes = [[0,0,1,1]];
labels = [0];
num_objs = 1;
boxes = torch.as_tensor(boxes, dtype=torch.float32)
labels = torch.as_tensor(labels, dtype=torch.int64)
image_id = torch.tensor([idx])
area = (boxes[:, 3] - boxes[:, 1]) * (boxes[:, 2] - boxes[:, 0])
# suppose all instances are not crowd
iscrowd = torch.zeros((num_objs,), dtype=torch.int64)
target = {}
target["boxes"] = boxes
target["labels"] = labels
target["image_id"] = image_id
target["area"] = area
target["iscrowd"] = iscrowd
if self.transforms is not None:
img, target = self.transforms(img, target)
return img, target
def __len__(self):
return len(self.imgs)
我的主方法如下所示，
def main():
# train on the GPU or on the CPU, if a GPU is not available
device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')
# our dataset has 16 classes - background and others
num_classes = 16
# use our dataset and defined transformations
dataset = SomeDataset('trainImages', get_transform(train=True))
print(f"{len(dataset)} number of images in training dataset")
dataset_validation = SomeDataset('valImages', get_transform(train=True))
print(f"{len(dataset_validation)} number of images in validation dataset")
# define training and validation data loaders
data_loader = torch.utils.data.DataLoader(
dataset, batch_size=20, shuffle=True, num_workers=4,
collate_fn=utils.collate_fn)
data_loader_val = torch.utils.data.DataLoader(
dataset_validation, batch_size=10, shuffle=False, num_workers=4,
collate_fn=utils.collate_fn)
# get the model using our helper function
#model = get_model_instance_segmentation(num_classes)
model = get_rcnn(num_classes);
# move model to the right device
model.to(device)
# construct an optimizer
params = [p for p in model.parameters() if p.requires_grad]
#optimizer = torch.optim.SGD(params, lr=0.005, momentum=0.9, weight_decay=0.0005);
optimizer = torch.optim.Adam(params, lr=0.0005);
# and a learning rate scheduler
lr_scheduler = torch.optim.lr_scheduler.StepLR(optimizer,
step_size=3,
gamma=0.1)
# let's train it for 10 epochs
num_epochs = 5
for epoch in range(num_epochs):
# train for one epoch, printing every 10 iterations
train_one_epoch(model, optimizer, data_loader, device, epoch, print_freq=100)
# update the learning rate
lr_scheduler.step()
# evaluate on the test dataset
#evaluate(model, data_loader_test, device=device)
print("That's it!")
return model;
当我运行我的代码时，它会运行几个数据（例如其中的10个），然后停止并输出此错误。
IndexError: Caught IndexError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "/usr/local/lib/python3.6/dist-packages/torch/utils/data/_utils/worker.py", line 178, in _worker_loop
data = fetcher.fetch(index)
File "/usr/local/lib/python3.6/dist-packages/torch/utils/data/_utils/fetch.py", line 44, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/usr/local/lib/python3.6/dist-packages/torch/utils/data/_utils/fetch.py", line 44, in <listcomp>
data = [self.dataset[idx] for idx in possibly_batched_index]
File "<ipython-input-114-e0ccd94603fd>", line 31, in __getitem__
xs = label_data[:,0:8:2];
IndexError: too many indices for array
错误发生在
model = main()到train_one_epoch()之间，且一直持续下去。我不明白为什么会这样。同时，请注意这只是数据集的一个实例。(<PIL.Image.Image image mode=RGB size=1024x1024 at 0x7F46FC0A94A8>, {'boxes': tensor([[ 628., 6., 644., 26.],
[ 633., 50., 650., 65.],
[ 620., 27., 637., 44.],
[ 424., 193., 442., 207.],
[ 474., 188., 496., 204.],
[ 383., 226., 398., 236.],
[ 399., 218., 418., 231.],
[ 42., 189., 63., 203.],
[ 106., 159., 129., 169.],
[ 273., 17., 287., 34.],
[ 225., 961., 234., 980.],
[ 220., 1004., 230., 1024.]]), 'labels': tensor([5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5]), 'image_id': tensor([0]), 'area': tensor([320., 255., 289., 252., 352., 150., 247., 294., 230., 238., 171., 200.]), 'iscrowd': tensor([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0])})