Grounded Segment Anything: From Objects to Parts

In this repo, we expand Segment Anything Model (SAM) to support text prompt input. The text prompt could be object-level:full_moon: (eg, dog) and part-level:last_quarter_moon: (eg, dog head). Furthermore，we build a Visual ChatGPT-based dialogue system 🤖💬 that flexibly calls various segmentation models when receiving instructions in the form of natural language.

[Blog] [Chinese Blog]

News

• 2023/04/14: Edit anything at more fine-grained part-level.
• 2023/04/11: Initial code release.

🚀New🚀 Edit on Part-Level

Part Prompt: "dog body"; Edit Prompt: "zebra" Part Prompt: "cat head"; Edit Prompt: "tiger" Part Prompt: "chair seat"; Edit Prompt: "cholocate" Part Prompt: "person head"; Edit Prompt: "combover hairstyle"

✨✨ Highlights ✨✨

Beyond class-agnostic mask segmentation, this repo contains:

• Grounded segment anything at both object level and part level.
• Interacting with models in the form of natural language.

These abilities come from a series of models, including:

Model	Function
Segment Anything	Segment anything from prompt
GLIP	Grounded language-image pre-training
Visual ChatGPT	Connects ChatGPT and segmentation foundation models
⭐VLPart⭐	Going denser with open-vocabulary part segmentation

FAQ

Q: When will VLPart paper be released ?

A: VLPart paper has been released. 🚀🚀🚀

Q: What is the difference between Grounded SAM and this project ?

A: Grounded SAM is Grounded DINO + SAM, and this project is GLIP/VLPart + SAM. We believe any open-vocabulary (text prompt) object detection model can be used to combine with SAM.

Usage

Install

See installation instructions.

Edit

python demo_part_edit.py

🤖💬 Integration with Visual ChatGPT

# prepare your private OpenAI key (for Linux)
export OPENAI_API_KEY={Your_Private_Openai_Key}
python chatbot.py --load "ImageCaptioning_cuda:0, SegmentAnything_cuda:1, PartPromptSegmentAnything_cuda:1, ObjectPromptSegmentAnything_cuda:0"

🌗 Prompt Segment Anything at Part Level

wget https://github.com/Cheems-Seminar/grounded-segment-any-parts/releases/download/v1.0/swinbase_part_0a0000.pth
wget https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth

python demo_vlpart_sam.py --input_image assets/twodogs.jpeg --output_dir outputs_demo --text_prompt "dog head"

Result:

🌕 Prompt Segment Anything at Object Level

wget https://github.com/Cheems-Seminar/grounded-segment-any-parts/releases/download/v1.0/glip_large.pth

python demo_glip_sam.py --input_image assets/demo2.jpeg --output_dir outputs_demo --text_prompt "frog"

Result:

🍭 Multi-Prompt

For multiple prompts, seperate each prompt with ., for example, --text_prompt "dog head. dog nose"

Model Checkpoints

• ViT-H SAM model
• VLPart Swin-Base
• GLIP Swin-Large

License

This project is under the CC-BY-NC 4.0 license. See LICENSE for details.

Acknowledgement

A large part of the code is borrowed from segment-anything, EditAnything, CLIP, GLIP, Grounded-Segment-Anything, Visual ChatGPT. Many thanks for their wonderful works.

Citation

If you find this project helpful for your research, please consider citing the following BibTeX entry.

@misc{segrec2023,
  title =        {Grounded Segment Anything: From Objects to Parts},
  author =       {Sun, Peize and Chen, Shoufa and Luo, Ping},
  howpublished = {\url{https://github.com/Cheems-Seminar/grounded-segment-any-parts}},
  year =         {2023}
}

@article{vlpart2023,
  title   =  {Going Denser with Open-Vocabulary Part Segmentation},
  author  =  {Sun, Peize and Chen, Shoufa and Zhu, Chenchen and Xiao, Fanyi and Luo, Ping and Xie, Saining and Yan, Zhicheng},
  journal =  {arXiv preprint arXiv:2305.11173},
  year    =  {2023}
}