YOLOv6训练自定义数据集报错：AssertionError无效目录路径

问题背景 #

严格遵循YOLOv6官方GitHub教程训练自定义数据集，数据集存放于YOLOv6父目录下，文件夹结构符合要求，训练/验证/测试图像齐全，但执行训练命令时持续报错。

报错信息 #

Using 1 GPU for training... 
training args are: Namespace(data_path='A/data.yaml', conf_file='configs/yolov6s.py', img_size=640, batch_size=16, epochs=50, workers=8, device='0', eval_interval=20, eval_final_only=False, heavy_eval_range=50, check_images=False, check_labels=False, output_dir='./runs/train', name='yolov6s-custom', dist_url='env://', gpu_count=0, local_rank=-1, resume=False, write_trainbatch_tb=False, stop_aug_last_n_epoch=15, save_ckpt_on_last_n_epoch=-1, distill=False, distill_feat=False, quant=False, calib=False, teacher_model_path=None, temperature=20, fuse_ab=False, bs_per_gpu=32, rank=-1, world_size=1, save_dir='runs/train/yolov6s-custom3')

Traceback (most recent call last):
  File "/content/gdrive/MyDrive/YOLO_Custom/YOLOv6/tools/train.py", line 128, in <module>
    main(args)
  File "/content/gdrive/MyDrive/YOLO_Custom/YOLOv6/tools/train.py", line 113, in main
    trainer = Trainer(args, cfg, device)
  File "/content/gdrive/MyDrive/YOLO_Custom/YOLOv6/yolov6/core/engine.py", line 54, in __init__
    self.train_loader, self.val_loader = self.get_data_loader(args, cfg, self.data_dict)
  File "/content/gdrive/MyDrive/YOLO_Custom/YOLOv6/yolov6/core/engine.py", line 361, in get_data_loader
    train_loader = create_dataloader(train_path, args.img_size, args.batch_size // args.world_size, grid_size,
  File "/content/gdrive/MyDrive/YOLO_Custom/YOLOv6/yolov6/data/data_load.py", line 41, in create_dataloader
    dataset = TrainValDataset(
  File "/content/gdrive/MyDrive/YOLO_Custom/YOLOv6/yolov6/data/datasets.py", line 68, in __init__
    self.img_paths, self.labels = self.get_imgs_labels(self.img_dir)
  File "/content/gdrive/MyDrive/YOLO_Custom/YOLOv6/yolov6/data/datasets.py", line 220, in get_imgs_labels
    assert osp.exists(img_dir), f"{img_dir} is an invalid directory path!"
AssertionError: ./images/train is an invalid directory path!

执行的训练命令 #

!python tools/train.py --batch-size 16 --conf configs/yolov6s.py --data data/data.yaml --device 0 --workers 8 --epochs 50 --img-size 640 --name yolov6s-custom

环境与配置说明 #

• data.yaml文件位于YOLOv6父目录的data文件夹中
• 数据集文件夹名为A，存放于YOLOv6父目录下
• data.yaml内容：

  train: ./images/train
  val: ./images/val
  test: ./images/test
  nc: 1
  names: ['xxx']
  

报错原因分析 #

报错核心是程序找不到./images/train路径，原因如下：

1. 相对路径基准错误：YOLOv6训练时的工作目录是YOLOv6根目录，data.yaml中的相对路径基于该目录，因此./images/train指向的是YOLOv6根目录下的images文件夹，而非父目录下数据集A内的对应路径。
2. data.yaml路径配置错误：未正确指向父目录下数据集A的图像文件夹。

解决步骤 #

方法1：使用绝对路径配置data.yaml #

直接写入数据集的绝对路径，避免相对路径混淆：

train: /content/gdrive/MyDrive/YOLO_Custom/A/images/train
val: /content/gdrive/MyDrive/YOLO_Custom/A/images/val
test: /content/gdrive/MyDrive/YOLO_Custom/A/images/test
nc: 1
names: ['xxx']

方法2：修正相对路径 #

利用../返回父目录，指向数据集A的对应路径（基于YOLOv6根目录）：

train: ../A/images/train
val: ../A/images/val
test: ../A/images/test
nc: 1
names: ['xxx']

方法3：调整数据集位置 #

将数据集文件夹A移动到YOLOv6根目录下，保持data.yaml原有路径配置不变。

验证方法 #

修改完成后，在YOLOv6根目录下执行以下命令测试路径有效性：

ls ../A/images/train

若能列出文件夹内文件，说明路径配置正确，重新执行训练命令即可。

内容的提问来源于stack exchange，提问作者tbensap18