v2.5.0

CVHub520/X-AnyLabelingv2.5.0Oct 15, 2024by CVHub520

AI Summary

X-AnyLabeling v2.5.0 is a major release introducing significant AI model integrations including SAM2.1, Florence-2, UPN, YOLOv5-sahi, and DocLayout-YOLO models. The release also adds powerful interactive visual-text prompting capabilities for generic vision tasks and improves data handling with COCO keypoint annotations support.

Key Highlights

Added support for SAM2.1 (Segment Anything Model 2.1) for advanced image segmentation
Introduced Florence-2 model support for enhanced vision-language tasks
New interactive visual-text prompting for generic vision tasks
Added COCO keypoint annotations import/export capability
Integrated UPN model for object proposal generation

New Features

SAM2.1 model support for image segmentation
Florence-2 model integration
Interactive visual-text prompting for vision tasks
UPN model for proposal box generation
YOLOv5-sahi model support
DocLayout-YOLO model support
Range selection for label batch modification
Export options dialog with additional path selection
COCO keypoint annotations import/export
Bounding box coloring by category or instance
Loop through each label action
Rectangle mode optimization to use minimum bounding box
Multiple bug fixes for import, export, UI, and model preprocessing

Full Release Notes

## Released versions

| Version | MD5                                |
|---------|------------------------------------|
| X-AnyLabeling-CPU.exe | e92cbfc03300c40312f5b4d4aa821200 |
| X-Anylabeling-Linux-CPU | 34f765b5cba2cd6a58ce05931045958a |

> [!NOTE]  
> For GPU acceleration and the macOS version, please consult the guide at  [this link](https://github.com/CVHub520/X-AnyLabeling/blob/main/docs/en/get_started.md).  
> If you need to use the following advanced features, please refer to the corresponding documentation to install additional dependencies.
> 
> 1. Video object tracking based on Segment-Anything-2: [Installation Guide](../../examples/interactive_video_object_segmentation/README.md)
> 2. Object proposal generation based on UPN: [Installation Guide](../../examples/detection/hbb/README.md)
> 3. Interactive visual-text prompting for generic vision tasks: [Installation Guide](../../examples/detection/hbb/README.md)

[![Open Vision](https://github.com/user-attachments/assets/b2c1419b-540b-44fb-988e-a48572268df7)](https://www.youtube.com/watch?v=QtoVMiTwXqk)
<div align="center"><strong>Interactive Visual-Text Prompting for Generic Vision Tasks</strong></div>
</br>

## Important updates
- 🚀 feat(model): add support interactive visual-text prompting for generic vision tasks by @CVHub520 in 8f5768bf56dae94e0d300ed008a4ee0496e159fa
- 🚀 feat(auto-labeling): optimize rectangle mode to use minimum bounding box by @CVHub520 in 4118efdc37987134344fd10e5010113ed36f41e4
- 🚀 feat(model): add support for SAM2.1 model by @CVHub520 in 812634da7c995d9360830445ec3e41d668e754b5
- 🚀 feat(model): add support for Florence-2 model (#679) by @CVHub520 in 67aac31c748c5cce576e4fc9c58958d23c59e58e 
- 🚀 feat(model): add support for the upn model to generate proposal boxes by @CVHub520 in 05eea0a00ad6d79ce1782a1367dc09046e12eed6
- 🚀 feat(model): add support for yolov5-sahi model by @CVHub520 in 455b3549c6366d390cacced7cbebcb1fb6fadbe5
- 🚀 feat(tool): add range selection for label batch modification (#708) by @CVHub520 in dcb7c510ff71116659056678f814ff497bb9a5f2
- 🚀 feat(export): add options dialog with additional export path selection (#702) by @CVHub520 in 398cb6bc65e05b53da6c004e6ea40559f699af4e
- 🚀 feat(import/export): add support for importing/exporting COCO keypoint annotations (#190) by @CVHub520 in 43cd1da21058a4c83790fb4c36126b9faab6e347
- 🚀 feat(model): Added support for DocLayout-YOLO model
- 🚀 feat(drawer): add option to color bounding boxes by category or instance by @CVHub520 in ba719c553e9af247d54cf6f0f6377913b89b9eff
- 🚀 feat(ui): add action to loop through each label by @julianstirling in 96f0617a3ff3e164131a98b2bd4bf56de214ab07
- 🐛 fix(import): handle invalid file paths in natural sort (#734) by @CVHub520 in 2b27a90c34ff0b8dbdd059abc39266835f46d43f
- 🐛 fix(export): improve mask overlapping handling in custom_to_mask method by @CVHub520 in 2dd1968f39e3c5bf8e56b4517f4ad27c1b86eca2
- 🐛 fix(tool): fix the path parsing error of the save_crop function by @wpNZC in 3a7c1e2e92b6121376d86f0a743a4dfc12a72607
- 🐛 fix(ui): disable delete action when no shapes present by @CVHub520 in 71d4cf2b5413e4fa122492ee1ef37d50dd7ea9a1
- 🐛 fix(model): correct image normalization in Recognize-Anything-Model preprocessing (#657) by @CVHub520 in 0b07dc4e0e54b66762053bc9c221aaabaece842e
- 📚 docs(installation): add onnxruntime compatibility by @CVHub520 in 37e87a2a6014537bb9fa9d00be822575d21fca11
- 🎨 refactor(ui): modernize GroupIDModifyDialog with improved styling by @CVHub520 in 7545561eea381cc1da8daf18feda2399f92f126d

**Full Changelog**: https://github.com/CVHub520/X-AnyLabeling/compare/v2.4.4...v2.5.0