[Revised Edition] Deep Learning Computer Vision Complete Guide
This course will help you become a deep learning-based computer vision expert needed in the field through in-depth theoretical explanations of Object Detection and Segmentation, along with practical examples that can be immediately applied in real-world work environments.
4,014 learners
Level Intermediate
Course period Unlimited
Notice on solution to error when performing train_detector using mmdetection
hello,
ย
Due to recent changes in the mmdetection package, when executing train_detector(model, datasets, cfg, distributed=False, validate=True), an error similar to 'ConfigDict' object has no attribute 'device' occurs.
ย
You can check the issue at https://github.com/open-mmlab/mmdetection/issues/7901.
ย
To solve this, add cfg.device='cuda' to the config settings as shown below.
ย
For example, in the cfg settings in the practice code, you can add cfg.device='cuda' at the very end as shown below.
ย
I'm not sure if this is a bug or intentional, so I won't modify the practice code on github for now.
ย
If the problem still occurs after observing for about a week, I will revise the practice code on github and notify you.
ย
thank you
ย
# Modify environment parameters for dataset.
cfg.dataset_type = 'PetDataset'
cfg.data_root = '/content/data/'
ย
# Modify environment parameters type, data_root, ann_file, and img_prefix for train, val, and test datasets.
cfg.data.train.type = 'PetDataset'
cfg.data.train.data_root = '/content/data/'
cfg.data.train.ann_file = 'train.txt'
cfg.data.train.img_prefix = 'images'
ย
.....other config settings
ย
# Settings to prevent 'ConfigDict' object has no attribute 'device' error.
cfg.device='cuda'