Skip to content

Latest commit

 

History

History
76 lines (68 loc) · 4 KB

urpc_2020.md

File metadata and controls

76 lines (68 loc) · 4 KB

URPC2020

@inproceedings{liu2021dataset,
  title={A dataset and benchmark of underwater object detection for robot picking},
  author={Liu, Chongwei and Li, Haojie and Wang, Shuchang and Zhu, Ming and Wang, Dong and Fan, Xin and Wang, Zhihui},
  booktitle={2021 IEEE International Conference on Multimedia \& Expo Workshops (ICMEW)},
  pages={1--6},
  year={2021},
  organization={IEEE}
}

The dataset contains 5,543 underwater images for training, 800 and 1,200 underwater images for testing (test-A and test-B set), covering four categories: holothurian, echinus, scallop, and starfish.

Download URPC2020 Dataset

The Underwater Robot Professional Contest (URPC) 2020 dataset, including training set, test-A set, and test-B set from here. You can also download the processed data from here.

For validation, we randomly divides the URPC2020 training set into training and validation groups with 4,434 and 1,019 images, respectively. If users want to divide by their own, tools/misc/write_txt.py should be used to split the train and val set first. Then tools/dataset_converters/xml_to_json.py can use to convert xml style annotations to coco format.

The data structure is as follows:

lqit
├── lqit
├── tools
├── configs
├── data
│   ├── URPC
│   │   ├── annotations_json       # coco style annotations
│   │   │   ├── train.json         # training group from training set, with 4,434 images
│   │   │   ├── val.json           # validation group from training set, with 1,019 images
│   │   │   ├── train_all.json     # training set, with all 5,543 images
│   │   │   ├── test-A.json        # testing-A set, with 800 images
│   │   │   ├── test-B.json        # testing-B set, with 1,200 images
│   │   │   ├── train-image        # training images
│   │   │   │   ├── 000001.jpg
│   │   │   │   ├── 000002.jpg
│   │   │   │   ├── ...
│   │   │   ├── test-A-image       # test-A images
│   │   │   │   ├── 000001.jpg
│   │   │   │   ├── 000002.jpg
│   │   │   │   ├── ...
│   │   │   ├── test-B-image       # test-B images
│   │   │   │   ├── 000001.jpg
│   │   │   │   ├── 000002.jpg
│   │   │   │   ├── ...
|   |   ├── source_data            # source data download from https://openi.pcl.ac.cn/OpenOrcinus_orca/URPC_opticalimage_dataset/datasets
│   │   │   ├── ImageSets          # get training, vaidation, testing image name from scripts
│   │   │   │   ├── train.txt
│   │   │   │   ├── val.txt
│   │   │   │   ├── train_all.txt
│   │   │   │   ├── test-A.txt
│   │   │   │   ├── test-B.txt
│   │   │   ├── ImageMetas         # get image meta information from scripts
│   │   │   │   ├── train-image-metas.pkl
│   │   │   │   ├── val-image-metas.pkl
│   │   │   │   ├── train_all-image-metas.pkl
│   │   │   │   ├── test-A-image-metas.pkl
│   │   │   │   ├── test-B-image-metas.pkl
│   │   │   ├── train-box          # pascal voc style annotations for the training set
│   │   │   │   ├── 000001.xml
│   │   │   │   ├── 000002.xml
│   │   │   │   ├── ...
│   │   │   ├── test-A-box         # pascal voc style annotations for the test-A set
│   │   │   │   ├── 000001.xml
│   │   │   │   ├── 000002.xml
│   │   │   │   ├── ...
│   │   │   ├── test-B-box         # pascal voc style annotations for the test-B set
│   │   │   │   ├── 000001.xml
│   │   │   │   ├── 000002.xml
│   │   │   │   ├── ...