自制数据集训练报错Error occurred during training with self-made dataset

[gobranchya]

我在用自己的数据集跑YOWOv2的算法，出现了以下报错，不知道如何解决：
File "train.py", line 330, in
train()
File "train.py", line 222, in train
for iter_i, (frame_ids, video_clips, targets) in enumerate(dataloader):
File "/root/miniconda3/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 521, in next
data = self._next_data()
File "/root/miniconda3/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1203, in _next_data
return self._process_data(data)
File "/root/miniconda3/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1229, in _process_data
data.reraise()
File "/root/miniconda3/lib/python3.8/site-packages/torch/_utils.py", line 425, in reraise
raise self.exc_type(msg)
IndexError: Caught IndexError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "/root/miniconda3/lib/python3.8/site-packages/torch/utils/data/_utils/worker.py", line 287, in _worker_loop
data = fetcher.fetch(index)
File "/root/miniconda3/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 44, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/root/miniconda3/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 44, in
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/root/YOWOv2/dataset/ava.py", line 156, in getitem
frame_idx, video_clip, target = self.pull_item(idx)
File "/root/YOWOv2/dataset/ava.py", line 178, in pull_item
keyframe_info = self._image_paths[video_idx][frame_idx - 1]
IndexError: list index out of range

Try to modify some specific parameters to be relevant to your own dataset. I am using a 60-second video with 30 frames per second to detect 6 actions, and modify the following modules:
尝试将一些特定的参数修改为自己数据集相关，我使用的60秒30帧的视频检测6个动作，对以下几个模块进行修改：
class AVA_Dataset(Dataset):
def init(self,
cfg,
data_root,
is_train=False,
img_size=224,
transform=None,
len_clip=60,
sampling_rate=1):
self.num_classes = 6

if name == 'main':
import cv2
from transforms import Augmentation, BaseTransform

is_train = False
img_size = 224
len_clip = 60
sampling_rate = 1

parser.add_argument('-K', '--len_clip', default=60, type=int,
help='video clip length.')

I know the problem lies in the following function, but I don't know how to modify it.
我知道问题出现在以下函数，不知道如何修改：
def pull_item(self, idx):
# Get the frame idxs for current clip. We can use it as center or latest
video_idx, sec_idx, sec, frame_idx = self._keyframe_indices[idx]
clip_label_list = self._keyframe_boxes_and_labels[video_idx][sec_idx]

    # check label list
    assert len(clip_label_list) > 0
    assert len(clip_label_list) <= self._max_objs

    # get a sequence
    seq = self.get_sequence(
        frame_idx,
        self.seq_len // 2,
        self.sampling_rate,
        num_frames=len(self._image_paths[video_idx]),
    )
    image_paths = [self._image_paths[video_idx][frame - 1] for frame in seq]
    keyframe_info = self._image_paths[video_idx][frame_idx - 1]

    # load a video clip
    video_clip = []
    for img_path in image_paths:
        frame = Image.open(img_path).convert('RGB')
        video_clip.append(frame)
    ow, oh = frame.width, frame.height

    # Get boxes and labels for current clip.
    boxes = []
    labels = []
    for box_labels in clip_label_list:
        bbox = box_labels[0]
        label = box_labels[1]
        multi_hot_label = np.zeros(1 + self.num_classes)
        multi_hot_label[..., label] = 1.0

        boxes.append(bbox)
        labels.append(multi_hot_label[..., 1:].tolist())

    boxes = np.array(boxes).reshape(-1, 4)
    # renormalize bbox
    boxes[..., [0, 2]] *= ow
    boxes[..., [1, 3]] *= oh
    labels = np.array(labels).reshape(-1, self.num_classes)

    # target: [N, 4 + C]
    target = np.concatenate([boxes, labels], axis=-1)

    # transform
    video_clip, target = self.transform(video_clip, target)
    # List [T, 3, H, W] -> [3, T, H, W]
    video_clip = torch.stack(video_clip, dim=1)

    # reformat target
    target = {
        'boxes': target[:, :4].float(),  # [N, 4]
        'labels': target[:, 4:].long(),  # [N, C]
        'orig_size': [ow, oh],
        'video_idx': video_idx,
        'sec': sec,

    }

    return [video_idx, sec], video_clip, target