Starstruk Logo MarkStarstruk

v2.4.0

CVHub520/X-AnyLabelingv2.4.0Jul 14, 2024by CVHub520

AI Summary

X-AnyLabeling v2.4.0 is a major release featuring new AI model support including SAM2, GroundingSAM2, RT-DETRv2, RAM++, YOLOW-RAM++, and Depth-Anything v2, along with enhanced annotation capabilities for tracking, KIE linking, and various import/export formats. The update also includes UI improvements like continuous drawing mode, shape union, crosshair customization, and a refreshed application icon.

Key Highlights

  • Added support for Segment-Anything-2 (SAM2) and GroundingSAM2 models
  • New tracking support for hbb/obb/seg/pose tasks with Bot-Sort and ByteTrack
  • Implemented Depth-Anything v2 model with color and grayscale rendering modes
  • Added masked image saving functionality and enhanced annotation import/export (PPOCR-KIE, yolo-pose, voc-format)
  • UI enhancements: continuous drawing mode toggle, shape union, crosshair customization, and confidence score display

New Features

  • Masked image saving functionality
  • Tracking support for hbb/obb/seg/pose tasks
  • GroundingSAM2 model support
  • Lightweight Japanese recognition model
  • Segment-Anything-2 (SAM2) model
  • Move mode parameter in label_dialog
  • System clipboard integration (copy mode)
  • ODVG annotation import/export for Grounding DINO
  • RT-DETRv2 model support
  • RAM++ and YOLOW-RAM++ image captioning models
  • KIE linking lines with arrowheads
  • Shape-level information display and export
  • Depth-Anything model with color/grayscale modes
  • PPOCR-KIE annotation import/export
  • KIE linking field annotation
  • Out-of-pixmap rotation shapes annotation
  • Depth-Anything v2 model support
  • PPOCR label import/export
  • Continuous drawing mode toggle
  • Union of multiple selected rectangle shapes
  • Delete labels by checkbox selection
  • Open previous/next labeled image
  • Crosshair and marking box style customization
  • Polygon to hbb conversion widget
  • YOLO-pose import/export
  • VOC-format export for polygon shapes
  • Auto-labeling with preserve existing annotations option
  • Real-time confidence parameter adjustments
  • Keypoint visibility feature
  • YOLOv8-world and YOLOv8-oiv7 support
  • Confidence score display

Full Release Notes

## Released versions

| Version | MD5                                |
|---------|------------------------------------|
| X-AnyLabeling-CPU.exe | 8442c84cafdf6deddfe9e5fceb9af17e |
| X-Anylabeling-Linux-CPU | b13829a34a1ff0f0b193c51fc4d951ea |

> [!NOTE]  
> To utilize GPU acceleration or the macOS version, refer to the guide at [this](https://github.com/CVHub520/X-AnyLabeling/blob/main/docs/en/get_started.md).

## Supported models

| Task                     | Model                  |
|--------------------------|------------------------|
| **Image Classification** | ResNet50, InternImage, YOLOv5-cls, YOLOv8-cls, PULC Person/Vehicle Attribute |
| **Object Detection**     | YOLOv5, YOLOv6, YOLOv7, YOLOv8, YOLOv9, YOLOv10, YOLOX, YOLO-NAS, DAMO_YOLO, GOLD_YOLO, RT-DETR, RTMDet |
| **Instance Segmentation** | YOLOv5-seg, YOLOv8-seg |
| **Keypoint Detection**   | RTMPose, DWPose, YOLOv6-face, YOLOv8-pose |
| **Oriented Object Detection** | YOLOv5_obb, YOLOv8_obb |
| **Multi-Object Tracking** | Bot-Sort, ByteTrack   |
| **Segment Anything**     | SAM2, SAM, SAM-HQ, EdgeSAM, MobileSAM, EfficientViT-SAM, Med-SAM2D |
| **Optical Character Recognition**                  | PPOCRv4                |
| **Land Detection**       | CLRNet                 |
| **Image Captioning**     | RAM++, RAM                   |
| **Visual Language Model** | Grounding-DINO, Chinese-CLIP, YOLO-World    |
| **Depth Estimation** | Depth Anything v2, Depth Anything |

<img src="https://github.com/user-attachments/assets/c65db18f-167b-49e8-bea3-fcf4b43a8ffd" width="100%" />

## Important updates

* Bumped to v2.4.0 with a refreshed X-Anything app icon by @CVHub520 in fd1d8e990dd65e001ce3d34d1f05a8ec3bbfd42f
* Implement structured ISSUE_TEMPLATE for better issue management by @CVHub520  in 6c1a7b3ade8103d2d8bbbb4f5e77aac70e8a2af2
* Create SECURITY.md file by @CVHub520 in 838bc294b5f4163b6f7738281812b7e082e137c8
* Optimize code related to "actions" by @PairZhu in a377764bf02f6bd5d69949df10f56d2c7f9a99c4

## New features

* Implemented masked image saving functionality by @CVHub520 in f076e28db86e52d985f8d5dcfced4204c82651ab
* Add support to tracking by hbb/obb/seg/pose task  by @CVHub520 in 479bfce22ae70f3f51d337d5078c79cdf5b842a9
* Support GroundingSAM2 model by @CVHub520  in b242f15761821a739a2862b5f485179f67bb587b
* Support lightweight model for Japanese recognition by @CVHub520 in 1a6abf5d5787c3b6f8e7277f5e4e6462d385916f
* Support Segment-Anything-2 model by @CVHub520 in 4e6857215afabe3a4c76dcd30e027219e02f9de8
* Enable move_mode parameter in label_dialog for enhanced interaction control by @CVHub520 in 4a943ad5f36349a6b4a4229e5246974829fb9a09
* Add use_system_clipboard action by @PairZhu in d7c9430ff4c16a80b81b2a47c118b242d78d72f4
* Enable import/export of ODVG annotations to and from the Grounding DINO dataset by @CVHub520 in 918a8b57fa7f2198613b60c723603fc2cae127fe
* Add support to RT-DETRv2 model by @CVHub520 in 0fe5cff437c79a418e41c0622753245e9972ef2d
* Add support RAM++ and YOLOW-RAM++ model by @CVHub520 in c1aa103e3934618a507c1cb9f3a9a1d1911ae537
* Implement feature to draw KIE linking lines with arrowheads by @CVHub520 in b8f98f142863cb627c6ad6a2587d5546f8812553
* Add support for displaying and exporting shape-level information by @CVHub520 in b9f201ca102c7b779b69ab3aeba5f9c9ae8437da
* Enable Depth-Anything model prediction with color and grayscale rendering modes by @CVHub520 in 837a6425fc6e28ff320bfc56902d5188af2a8b6a
* Add import and export functionality for PPOCR-KIE annotations by @CVHub520 in bbab395eebc2585e507f8ff9eff3e067e5df8e97
* Add support to annotate KIE linking field by @CVHub520 in 4441704a76e719b118a01927cd49378a48d223ae
* Add support for annotating out-of-pixmap rotation shapes by @CVHub520 in fe3cb53c3c0c95a73f76de3a264991c9ab2742ab
* Add support for the depth-anything-v2 model by @CVHub520 in 6dd8ab1efba736f12a8fb0ec3541590a5062a09d
* Add import and export functionality for PPOCR label by @CVHub520 in ae79d5661a64bf492baaa2a5fd6de6eae000a856
* Add toggle for continuous drawing mode without automatic switch to edit mode by @CVHub520 in 7ba645870f570c61bd9554a7a455975f08f9cfa6
* Support union of multiple selected rectangle shapes by @CVHub520 in 8244cce94f0f515b32cb6996b6f76392fe1c5d08
* Add system_clipboard copy mode by @PairZhu in 24d14f2e52d337bc65dc8d31797dd8407a58d562
* Add ability to delete label items based on checkbox selection by @CVHub520 in 6020b855ed4e007712616fe5da057f9d24edfa23
* Enable opening previous/next labeled image by @CVHub520 in 8f5d0e678272a2c478d52f9e9d1f545f72582503
* Implement crosshair and marking box style customization by @CVHub520 in 233dd4ff2940f3482086187a5337944b9d183baf
* Add widget for converting polygon to hbb support by @CVHub520 in 00d3fffcd24da871041759f57f9bb4d9ae1f4206
* Add yolo-pose import and export functionality by @CVHub520 in 99fb6c87e0b0eda4d68a8338b3c8de9fbb67d30e
* Enable export voc-format annotations for polygon shape by @CVHub520 in 15ee242b1ea13ec86bc3ee8f3382a3d0feba5eab
* Add preserve existing annotations checkbox to auto-labeling settings and allow real-time confidence parameter adjustments on the interface by @CVHub520 in ec1ae44f7a51a77f4079ed9a4c9e2734c45bf9fe
* Add visibility feature for keypoint detection task by @CVHub520 in 5d8a7c1f35568321517104501b0432ede0c326fe
* Support yolov8-world and yolov8-oiv7 model by @CVHub520 in 98a95db7f7a310b1a5c094c912bc8a4aea2a86cd
* Add feature to display confidence score by @CVHub520 in e6de5e6f695f24f5c35ea4afac721a399364246c

## Docs

* Add [faq.md](https://github.com/CVHub520/X-AnyLabeling/blob/main/docs/zh_cn/faq.md) file by @CVHub520 in ea098d0a3e3e62e515a9ad7a6e0aa45a25cbd648
* Add OCR example by @CVHub520  in 37b3150a1ee83e7d973ae1dcd2e50bc03ee83d8c
* Add MOT example by @CVHub520 in 66d1a559de6424c1b308474b24eba2d704d38332
* Add Pose Estimation example by @CVHub520  in 9874bee83465a55351ebdd2fc59d1772d31a2887
* Add Segmentation example by @CVHub520 in b49bd6182e90f450301d55486f19a63e6f2dcb60
* Add HBB Object Detection example by @CVHub520  in e52c5013ce48d6626a1431511294c3aec9d75c11
* Add OBB Object Detection example by @CVHub520 in 34d6411fafa1c1ed58e800cff06a8e7c6d30c336
* Add Depth Estimation example by @CVHub520 in f650b9ad23bc370fa97d953dd31dc0d5047c7905
* Add Description examples by @CVHub520 in 106c1bf2c8091f091282dcd4e49fdfb451a06341
* Add Shape-level Classification examples by @CVHub520 in cf7ff2c6e9c4bdcd5fe49ffbe0d29b110691b314
* Add Image-level Classification examples by @CVHub520 in c8109c1a410ec456ed20d1cd9eaed318236f13a0

## Bug fixed

* Fixed image distortion issue during brightness and contrast adjustment by @CVHub520 in 252ded051fa5db29fe92eae090cad70c9c77a9c6
* Fixed type error in fillRect by converting float values to int for compatibility with higher Python version by @CVHub520 in 71cfc307e699a4a42864702714aced5cf0f45043
* Resolved too many values to unpack error during YOLO class post-process by @CVHub520 in a69077d07d5d72b509a74520f4cc882e4b507c7d
* Fixed invalid literal for int() with base issue by @CVHub520 in 9380142fa2c2e16a791cf33ac35a568bb3eb8445
* Avoided directory not empty error when loading model @CVHub520 in 599327b2bf249c68a2e89a6eb309e451e22afbc5
* Prevented crash when switching from image directory to imported image by @CVHub520 in 24022683f67a0f17783c5be922d6b6e445a80a17
* Fixed the issue where BMP image files could not be loaded due to the missing '_getexif' attribute by @UnlimitedWand in 6f6407789b261f7636defe1f34626f9df21f11a8

**Full Changelog**: https://github.com/CVHub520/X-AnyLabeling/compare/v2.3.7...v2.4.0