search

cxliu0 / OA-MIL

[ECCV 2022] Robust Object Detection With Inaccurate Bounding Boxes
MIT License
34 stars 3 forks source link

About annotation files #11

Open cxq1 opened 1 year ago

cxq1 commented 1 year ago

Hello, did you use coco2017_train.pkl or coco2017_train.json in the experiment? I found that using gen_noisy_cococ.py is to generate coco2017_train.pkl

UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 0: invalid start byte ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: 1) local_rank: 0 The above problem occurred when I used coco2017_train.pkl as the training annotation file

cxliu0 commented 1 year ago

We did use coco2017_train.pkl to train clean-FasterRCNN and clean-RetinaNet, and this file works fine.

We have not encountered this error before. Perhaps you were training the model in Windows system? If so, you may refer to this issue.

cxq1 commented 1 year ago

We did use coco2017_train.pkl to train clean-FasterRCNN and clean-RetinaNet, and this file works fine.

We have not encountered this error before. Perhaps you were training the model in Windows system? If so, you may refer to this issue.

I am running on Linux system, can you tell me the version of pycocotools used? I guess it has something to do with writing a function. Also please tell me the version of pytorch python if it is convenient.

cxliu0 commented 1 year ago

We just used MMDetection to process the COCO dataset instead of separately installing pycocotools.

In addition, the version of pytorch and python can be found here.