v3.2.4

CVHub520/X-AnyLabelingv3.2.4Sep 28, 2025by CVHub520

AI Summary

This release (v3.2.4) of X-AnyLabeling introduces significant enhancements including a dedicated multi-class image classifier with Ultralytics support, improved UI controls for label visibility and attribute selection, and cross-widget reference capabilities in VQA dialogs. The update also adds various quality-of-life improvements like keyboard navigation shortcuts, select/deselect all shapes, and better YOLO dataset handling.

Key Highlights

  • New multi-class image classifier with streamlined workflow and Ultralytics support
  • Checkboxes for description and label visibility control in the UI
  • Cross-widget reference support in VQA dialog using @widget_title syntax
  • Keyboard shortcuts (A/D) for image navigation
  • Select/deselect all shapes feature and loop select labels functionality

New Features

  • Dedicated multi-class image classifier with streamlined workflow (#480)
  • Ultralytics image classification task support
  • Delete group IDs from objects (#1141)
  • Checkboxes for description and label visibility control (#1139)
  • Loop select labels for sequential shape selection (#1138)
  • Drawing rectangle shapes outside canvas with auto-clipping (#1137)
  • Radio button support for attribute selection (#1135)
  • Option to skip empty label files when creating YOLO datasets
  • Option to preserve existing annotations when uploading YOLO labels (#1125)
  • Custom provider support for chatbot with enhanced model dropdown
  • Cross-widget reference support in VQA dialog using @widget_title syntax
  • Support for paths wrapped in quotes for models and data
  • Select/deselect all shapes feature (#1092)
  • Keyboard shortcuts for image navigation (A/D)

Full Release Notes

https://github.com/user-attachments/assets/0652adfb-48a4-4219-9b18-16ff5ce31be0

> Baidu Cloud: https://pan.baidu.com/s/1pgaw02inCvbEgOme9ajDJA?pwd=e528

### 🚀 New Features

- Introduce a dedicated multi-class image classifier with a streamlined workflow (#480)
- Add support for Ultralytics image classification tasks
- Add support for deleting group IDs from objects (#1141)
- Add checkboxes for description and label visibility control (#1139)
- Add loop select labels functionality for sequential shape selection (#1138)
- Add support for drawing rectangle shapes outside canvas with auto-clipping (#1137)
- Add radio button support for attribute selection (#1135)
- Add an option to skip empty label files when creating YOLO datasets and update the UI (#1131)
- Add an option to preserve existing annotations when uploading YOLO labels (#1125)
- Add custom provider support for the chatbot and enhance the model dropdown feature
- Add cross-widget reference support in VQA dialog using `@widget_title` syntax
- Add support for paths wrapped in quotes for models and data
- Add select/deselect all shapes feature (#1092)
- Implement keyboard shortcuts for image navigation (A/D)

### 🐛 Bug Fixes

- Resolve inconsistent attribute behavior after shape creation and switching (#1134)
- Ensure linestrip's vertex is drawn regardless of selection state (#1134)
- Resolve issue where CUDA device count returns 0 after model export in Ultralytics training (#1126)
- Fix Windows path separator error in `train_script.py`
- Fix inconsistent shape order when using existing shapes for recognition in PP-OCR
- Fix issue where group ID info was only updated after successful modification

### 🛠️ Improvements

- Auto-update the attributes panel after shape creation (#1134)
- Improve shape selection logic for point and line types to enhance user interaction (#1134)
- Enhance shape selection for partial re-recognition in PP-OCR (#1113)
- Enhance AI prompts in VQA with cross-component and annotation data references
- Preserve original shape properties when skipping detection in OCR (#1116)

### 🌟 Contributors

A total of 3 developers contributed to this release.

Thank @sckiyo, @Vlad188-1, @CVHub520

**Full Changelog**: https://github.com/CVHub520/X-AnyLabeling/compare/v3.2.3...v3.2.4