Hi! Thanks for the code!

I tried to use your config_file in segmentation dir, but got an error. It seems in config.py there is no "type" key. I'm not familiar with openmmlab. Could you help me figure it out? Thanks a lot!

Script:

config_file = "./segmentation/configs/swin/swin_l_upper_w_jitter_inference.py"
ckpt_file = "../../models/seg_swin_l_uvo_finetuned.pth"
model = init_segmentor(config_file, ckpt_file, device="cuda:0")

Log:

Traceback (most recent call last):
  File "infer.py", line 20, in <module>
    model = init_segmentor(config_file, ckpt_file, device="cuda:0")
  File "/home/xin/anaconda3/envs/openmmlab/lib/python3.7/site-packages/mmseg/apis/inference.py", line 32, in init_segmentor
    model = build_segmentor(config.model, test_cfg=config.get('test_cfg'))
  File "/home/xin/anaconda3/envs/openmmlab/lib/python3.7/site-packages/mmseg/models/builder.py", line 49, in build_segmentor
    cfg, default_args=dict(train_cfg=train_cfg, test_cfg=test_cfg))
  File "/home/xin/anaconda3/envs/openmmlab/lib/python3.7/site-packages/mmcv/utils/registry.py", line 212, in build
    return self.build_func(*args, **kwargs, registry=self)
  File "/home/xin/anaconda3/envs/openmmlab/lib/python3.7/site-packages/mmcv/cnn/builder.py", line 27, in build_model_from_cfg
    return build_from_cfg(cfg, registry, default_args)
  File "/home/xin/anaconda3/envs/openmmlab/lib/python3.7/site-packages/mmcv/utils/registry.py", line 25, in build_from_cfg
    '`cfg` or `default_args` must contain the key "type", '
KeyError: '`cfg` or `default_args` must contain the key "type", but got {\'pretrained\': None, \'backbone\': {\'pretrain_img_size\': 384, \'embed_dims\': 192, \'depths\': [2, 2, 18, 2], \'num_heads\': [6, 12, 24, 48], \'drop_path_rate\': 0.2, \'window_size\': 12}, \'decode_head\': {\'in_channels\': [192, 384, 768, 1536], \'num_classes\': 2, \'loss_decode\': {\'type\': \'CrossEntropyLoss\', \'use_sigmoid\': False, \'loss_weight\': 1.0}}, \'auxiliary_head\': {\'in_channels\': 768, \'num_classes\': 2, \'loss_decode\': {\'type\': \'CrossEntropyLoss\', \'use_sigmoid\': False, \'loss_weight\': 1.0}}, \'train_cfg\': None}\n{\'train_cfg\': None, \'test_cfg\': None}'

Hi！ Sorry to bother! The links to download weights failed. Can you offer model weights ? I'm seeking for object detection model. Thank you! BR! George.

Firstly, Thank You for your amazing work.

There’s a question I would like to ask regarding the UVO Dataset. I would like to work on the dense dataset. Through the download link in the homepage of the challenge, I could access ‘UVO_video_train_dense.json’, ‘UVO_video_val_dense.json’ and ‘UVO_video_test_dense.json’ that specify the video ids of the train, validation and test datasets. However, I’m unsure how I could obtain the videos. Could you please guide me on how I could obtain the original and annotated videos ?

Many thanks

Hi, @dulucas , In these two config files, many "mask annotations" are used, like in this:

oid_train= dict( type='RepeatDataset', times=1, dataset=dict( type=dataset_type, data_root='data/oid/', img_dir = 'images/', ann_dir = 'masks/', split = ['train_clean_v2.txt', ], pipeline=train_pipeline ) ) and : uvo_dense_val = dict( type='RepeatDataset', times=1, dataset=dict( type=dataset_type, data_root='data/uvo/', img_dir = 'images/dense_val/', ann_dir = 'masks/dense_val/', split = ['dense_val_list.txt', ], pipeline=train_pipeline ) )

And I tried the code like this(https://github.com/alicranck/coco2voc)(url) to generate the mask pngs, but there maybe something wrong with the generated masks, the training losses are unusual. Could you please provide the code or scripts that you used to generate the mask? or giving a link you are referred?

Hello! Many thanks for this nice GitHub repo. I am wondering on which data split did you evaluate your model. Whether it was UVO sparse test or UVO dense test? In the paper I saw the statement of Challenge final results on UVO-Sparse test dataset. But the testing script in Github code loaded test annotation from the dense split. I guess dense split's test set was what the challenge targeted?

this checkpoint(seg_swin_l_mixed_pretrained.pth) seem broken.

import torch
torch.load('seg_swin_l_mixed_pretrained.pth')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/chen/miniconda3/envs/sota/lib/python3.9/site-packages/torch/serialization.py", line 600, in load
    with _open_zipfile_reader(opened_file) as opened_zipfile:
  File "/home/chen/miniconda3/envs/sota/lib/python3.9/site-packages/torch/serialization.py", line 242, in __init__
    super(_open_zipfile_reader, self).__init__(torch._C.PyTorchFileReader(name_or_buffer))
RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory

First of all thank the author for sharing such an excellent work！ When browsing the code, I found that there is a loop call in the get_targets function, is this not a problem?