SegPrompt: Boosting Open-world Segmentation via Category-level Prompt Learning
In this work, we propose a novel training mechanism termed SegPrompt that uses category information to improve the model's class-agnostic segmentation ability for both...
Tags：Paper and LLMsInstance Segmentation Semantic SegmentationPricing Type
- Pricing Type: Free
- Price Range Start($):
GitHub Link
The GitHub link is https://github.com/aim-uofa/segprompt
Introduce
The repository “aim-uofa/SegPrompt” contains the official implementation of the ICCV 2023 paper titled “SegPrompt Boosting Open-World Segmentation via Category-level Prompt Learning.” The authors propose SegPrompt for improving open-world segmentation through category-level prompt learning. They introduce a new benchmark called LVIS-OW, which involves reorganizing COCO and LVIS datasets into Known-Seen-Unseen categories for better evaluating open-world models. The repository provides dataset preparation instructions, benchmark details, and evaluation scripts. Acknowledgments are given to related repositories like Mask2Former and Detectron2, and the paper encourages proper citation if the project is used.
In this work, we propose a novel training mechanism termed SegPrompt that uses category information to improve the model’s class-agnostic segmentation ability for both known and unknown categories.
Content
1Zhejiang University, 2The University of Adelaide, Please follow the instructions in Mask2Former Here we provide our proposed new benchmark LVIS-OW. First prepare COCO and LVIS dataset, place them under $DETECTRON2_DATASETS following Detectron2 The dataset structure is as follows: Or you can directly use the command to generate from the json file of COCO and LVIS. We thank the following repos for their great works: If you found this project useful for your paper, please kindly cite our paper.
Related
LAMA utilizes a reinforcement learning framework combined with a motion matching algorithm. Reinforcement learning helps the model make appropriate decisions in various scenarios, while motion matching algorithms ensure that synthesized actions match real human actions. In addition, LAMA also utilizes the motion editing framework of manifold learning to cover various possible changes in interactions and operations.
