AttributeError: 'NoneType' object has no attribute 'shape'

Checklist

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

Describe the bug Hello, the training of the last few days was very normal, and this error suddenly appeared today. The same data is not wrong on another machine.

Reproduction

1. My config

# model settings
model = dict(
    type='RetinaNet',
    pretrained='torchvision://resnet50',
    backbone=dict(
        type='ResNet',
        depth=50,
        num_stages=4,
        out_indices=(0, 1, 2, 3),
        frozen_stages=1,
        style='pytorch'),
    neck=dict(
        type='FPN',
        in_channels=[256, 512, 1024, 2048],
        out_channels=256,
        start_level=1,
        add_extra_convs=True,
        num_outs=5),
    bbox_head=dict(
        type='FreeAnchorRetinaHead',
        num_classes=6,
        in_channels=256,
        stacked_convs=4,
        feat_channels=256,
        octave_base_scale=2,
        scales_per_octave=3,
        anchor_ratios=[0.5, 1.0, 2.0],
        anchor_strides=[8, 16, 32, 64, 128],
        target_means=[.0, .0, .0, .0],
        target_stds=[0.1, 0.1, 0.2, 0.2],
        loss_bbox=dict(type='SmoothL1Loss', beta=0.11, loss_weight=0.75)))
# training and testing settings
train_cfg = dict(
    assigner=dict(
        type='MaxIoUAssigner',
        pos_iou_thr=0.5,
        neg_iou_thr=0.4,
        min_pos_iou=0,
        ignore_iof_thr=-1),
    allowed_border=-1,
    pos_weight=-1,
    debug=False)
test_cfg = dict(
    nms_pre=3000,
    min_bbox_size=0,
    score_thr=0.05,
    nms=dict(type='nms', iou_thr=0.5),
    max_per_img=1800)
# dataset settings
dataset_type = 'VOCDataset'
data_root = '/home/shelly/mmdetection/data/VOCdevkit/'
img_norm_cfg = dict(
    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
train_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='LoadAnnotations', with_bbox=True),
    dict(type='Resize', img_scale=(1024, 1024), keep_ratio=True),
    dict(type='RandomFlip', flip_ratio=0.5),
    dict(type='Normalize', **img_norm_cfg),
    dict(type='Pad', size_divisor=32),
    dict(type='DefaultFormatBundle'),
    dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels']),
]

val_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='LoadAnnotations', with_bbox=True),
    dict(type='Resize', img_scale=(1024, 1024), keep_ratio=True),
    dict(type='RandomFlip', flip_ratio=0.0),
    dict(type='Normalize', **img_norm_cfg),
    dict(type='Pad', size_divisor=32),
    dict(type='DefaultFormatBundle'),
    dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels']),
]

test_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(
        type='MultiScaleFlipAug',
        img_scale=(1024, 1024),
        flip=False,
        transforms=[
            dict(type='Resize', keep_ratio=True),
            dict(type='RandomFlip'),
            dict(type='Normalize', **img_norm_cfg),
            dict(type='Pad', size_divisor=32),
            dict(type='ImageToTensor', keys=['img']),
            dict(type='Collect', keys=['img']),
        ])
]
data = dict(
    imgs_per_gpu=2,
    workers_per_gpu=2,
    train=dict(
        type=dataset_type,
        ann_file=data_root + 'VOC2012/ImageSets/Main/train.txt',   
        img_prefix=data_root + 'VOC2012/',    
        pipeline=train_pipeline),
    val=dict(
        type=dataset_type,
        ann_file=data_root + 'VOC2012/ImageSets/Main/test.txt',
        img_prefix=data_root + 'VOC2012/',   
        pipeline=val_pipeline),
    test=dict(
        type=dataset_type,
        ann_file=data_root + 'VOC2012/ImageSets/Main/test.txt',
        img_prefix=data_root + 'VOC2012/',   
        pipeline=test_pipeline))
# optimizer
optimizer = dict(type='SGD', lr=0.00125, momentum=0.9, weight_decay=0.0001)
optimizer_config = dict(grad_clip=dict(max_norm=35, norm_type=2))
# learning policy
lr_config = dict(
    policy='step',
    warmup='linear',
    warmup_iters=500,
    warmup_ratio=1.0 / 3,
    step=[8, 11])
checkpoint_config = dict(interval=1)
# yapf:disable
log_config = dict(
    interval=50,
    hooks=[
        dict(type='TextLoggerHook'),
        # dict(type='TensorboardLoggerHook')
    ])
# yapf:enable
# runtime settings
total_epochs = 12
device_ids = range(8)
dist_params = dict(backend='nccl')
log_level = 'INFO'
work_dir = './work_dirs/retinanet_free_anchor_r50_fpn_1x'
load_from = None
resume_from = None
workflow = [('train', 4),('val', 1)]

Environment

• OS: ubuntu16.04
• GCC: 7.3.0
• PyTorch version:1.2.0
• How you installed PyTorch: conda
• GPU mode: 1080ti
• CUDA and CUDNN version: 9.0 7.3

Error traceback

2019-12-14 13:01:55,044 - INFO - Epoch [3][3050/3648]   lr: 0.00125, eta: 5:12:16, time: 0.306, data_time: 0.004, memory: 8879, positive_bag_loss: 1.3066, negative_bag_loss: 0.1516, loss: 1.4582
2019-12-14 13:02:10,510 - INFO - Epoch [3][3100/3648]   lr: 0.00125, eta: 5:11:08, time: 0.309, data_time: 0.004, memory: 8879, positive_bag_loss: 1.4534, negative_bag_loss: 0.1561, loss: 1.6096
2019-12-14 13:02:37,388 - INFO - Epoch [3][3150/3648]   lr: 0.00125, eta: 5:10:36, time: 0.538, data_time: 0.004, memory: 8879, positive_bag_loss: 1.3954, negative_bag_loss: 0.1542, loss: 1.5496
2019-12-14 13:02:52,779 - INFO - Epoch [3][3200/3648]   lr: 0.00125, eta: 5:09:29, time: 0.308, data_time: 0.004, memory: 8879, positive_bag_loss: 1.3824, negative_bag_loss: 0.1583, loss: 1.5406
libpng error: bad adaptive filter value
Traceback (most recent call last):

  File "/home/shelly/mmdetection/tools/train.py", line 113, in <module>
    main()

  File "/home/shelly/mmdetection/tools/train.py", line 109, in main
    logger=logger)

  File "/home/shelly/mmdetection/mmdet/apis/train.py", line 60, in train_detector
    _non_dist_train(model, dataset, cfg, validate=validate)

  File "/home/shelly/mmdetection/mmdet/apis/train.py", line 232, in _non_dist_train
    runner.run(data_loaders, cfg.workflow, cfg.total_epochs)

  File "/home/shelly/anaconda3/envs/py37/lib/python3.7/site-packages/mmcv/runner/runner.py", line 363, in run
    epoch_runner(data_loaders[i], **kwargs)

  File "/home/shelly/anaconda3/envs/py37/lib/python3.7/site-packages/mmcv/runner/runner.py", line 263, in train
    for i, data_batch in enumerate(data_loader):

  File "/home/shelly/anaconda3/envs/py37/lib/python3.7/site-packages/torch/utils/data/dataloader.py", line 819, in __next__
    return self._process_data(data)

  File "/home/shelly/anaconda3/envs/py37/lib/python3.7/site-packages/torch/utils/data/dataloader.py", line 846, in _process_data
    data.reraise()

  File "/home/shelly/anaconda3/envs/py37/lib/python3.7/site-packages/torch/_utils.py", line 369, in reraise
    raise self.exc_type(msg)

AttributeError: Caught AttributeError in DataLoader worker process 0.
Original Traceback (most recent call last):
  File "/home/shelly/anaconda3/envs/py37/lib/python3.7/site-packages/torch/utils/data/_utils/worker.py", line 178, in _worker_loop
    data = fetcher.fetch(index)
  File "/home/shelly/anaconda3/envs/py37/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py", line 44, in fetch
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/home/shelly/anaconda3/envs/py37/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py", line 44, in <listcomp>
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/home/shelly/mmdetection/mmdet/datasets/custom.py", line 128, in __getitem__
    data = self.prepare_train_img(idx)
  File "/home/shelly/mmdetection/mmdet/datasets/custom.py", line 141, in prepare_train_img
    return self.pipeline(results)
  File "/home/shelly/mmdetection/mmdet/datasets/pipelines/compose.py", line 24, in __call__
    data = t(data)
  File "/home/shelly/mmdetection/mmdet/datasets/pipelines/loading.py", line 30, in __call__
    results['img_shape'] = img.shape
AttributeError: 'NoneType' object has no attribute 'shape'

Bug fix If you have already identified the reason, you can provide the information here. If you are willing to create a PR to fix it, please also leave a comment here and that would be much appreciated!

[SirRacha]

Same issue here, and no way to find out which file exactly is broken (in 5M files)

I found the image this way, out side of the training code

`

FIND IMAGE THAT CANT BE OPENED

from PIL import Image pname = list/of/filepaths for it in pname: try: img = Image.open(it) img.verify() # to verify if its an img img.close() #to close img and free memory space except (IOError, SyntaxError) as e: print(‘Bad file:’, it) `

[dreamflasher]

Same issue here, and no way to find out which file exactly is broken (in 5M files)