This github repository is a complimentary resource for the Instance Detection Challenge. From python files and jupyter notebook files we provided, you can see how the challenge dataset is organized, how the evaluation is done, and how the required json file is generated for benchmarking on EvalAI. We also provide detailed comparison between two baselines in paper.
2025/06: Object Instance Detection Challenge will be held at Workshop on Visual Perception and Learning in an Open World, CVPR 2025
2024/12/08: 1st Object Instance Detection Workshop was held at ACCV 2024.
2024/06/18: Object Instance Detection Challenge was held at Workshop on Visual Perception and Learning in an Open World, CVPR 2024
You may want to install necessary packages. Run pip install -r requirements.txt in your python environment. Note that we only tested the code on python >= 3.8.
Note: Evaluation protocol will recognize easy and hard by image id automatically. If participants need to submit a mini-version, please ensure your submission contains both easy and hard test images.
The challenge server at EvalAI requires participants to upload a json or csv file that lists results on the test-set. We also provide examples of submission.json and submission_fasterrcnn_dev.csvhere.
Should you have technical questions, please create an issue here. If you have high-level questions, please contact Qianqian Shen ([email protected]) and Yunhan Zhao ([email protected]).