"RuntimeError: The training pipeline of the dataset wrapper always return None" while using albu and MultiImageMixDataset

[WangChen100]

Thanks for your error report and we appreciate it a lot.

Checklist

1. I have searched related issues but cannot get the expected help.
2. I have read the FAQ documentation but cannot get the expected help.
3. The bug has not been fixed in the latest version.

Describe the bug When I use Albumentation and Mosaic as data augmentation in MMDetectoin 3.3.0, RuntimeError was reported.

Reproduction

a part of cofig is as follow:

train_pipeline = [
    dict(type='LoadImageFromFile', backend_args=backend_args),
    dict(type='LoadAnnotations', with_bbox=True),
    dict(type='Mosaic',img_scale=(1280, 1280),pad_val=114.0,bbox_clip_border=False),
    dict(type='RandomAffine', scaling_ratio_range=(0.1, 2), border=(-640, -640)),
    dict(type='MixUp',img_scale=(1280, 1280),ratio_range=(0.8, 1.6),pad_val=114.0,bbox_clip_border=False),
    dict(type='RandomChoiceResize',
        scales=[(960, 960), (1024, 1024), (1088, 1088), (1152, 1152), (1216, 1216), (1280, 1280)],
        keep_ratio=True),
     dict(type='Albu',
        transforms=albu_train_transforms,
        bbox_params=dict(
            type='BboxParams',
            format='pascal_voc',
            label_fields=['gt_bboxes_labels', 'gt_ignore_flags'],
            min_visibility=0.0,
            filter_lost_elements=True),
        keymap={
            'img': 'image',
            # 'gt_masks': 'masks',
            'gt_bboxes': 'bboxes'
        },
        skip_img_without_anno=True),
    dict(type='RandomGrayscale', prob=0.1, keep_channels=True),
    dict(type='RandomFlip', prob=0.5, direction=['horizontal', 'vertical']),
    dict(type='PackDetInputs')
]
train_dataloader = dict(
    batch_size=8,
    num_workers=16,
    dataset = dict(
        _delete_=True,
        # use MultiImageMixDataset wrapper to support mosaic and mixup
        type='MultiImageMixDataset',
        dataset=dict(
            type='CocoDataset',
            data_root='data/Fisheye8K/',
            ann_file='train/train.json',
            data_prefix=dict(img='train/images/'),
            pipeline=[
                dict(type='LoadImageFromFile', backend_args=backend_args),
                dict(type='LoadAnnotations', with_bbox=True)
            ],
            filter_cfg=dict(filter_empty_gt=False, min_size=32),
            backend_args=backend_args),
        pipeline=train_pipeline)
)

Error traceback I find this error is preduced in line151 of dataset_wrappers.py. It is caused by that updated_results is always None.

What's more, I inspect function transform() of class Albu() in transforms.py，the size of gt_bboxes in the input results is 0:

I also find similiar issues: "TypeError: argument of type 'NoneType' is not iterable" when use mixup and alub augmentation together #7746 [Fix]fix the bug that mix_results may be None #7530

How could I solve this problem?

[aerotractjack]

did you ever find a solution?