image_background_remover

Fix

• Bounded memory use during mask compositing. removeBg/removeBgBytes previously held the full-resolution decoded source image through the entire RGBA decode/mask-apply/encode pipeline. For typical 12MP+ camera photos this could hold 2-3 full-resolution buffers in memory at once, causing out-of-memory crashes on release builds and mid/low-RAM devices (the model itself only ever used a 320x320 copy). Added BackgroundRemover.instance.maxWorkingDimension (default 1600) — source images with a longest side above this are downscaled, preserving aspect ratio, before mask compositing.

Full Changelog: v2.0.0...v2.0.0

🎉 Major Update: Migration to flutter_onnxruntime

Breaking Changes

• Migrated from onnxruntime to flutter_onnxruntime package for better maintenance and support
• dispose() method is now asynchronous internally (no user code changes required)

New Features

• Modular Architecture: Refactored codebase into organized modules for better maintainability
  • services/onnx_session_manager.dart - ONNX Runtime session management
  • utils/image_processor.dart - Image processing utilities
  • utils/mask_processor.dart - Mask processing utilities
  • utils/background_composer.dart - Background composition utilities
• Better Resource Management: Improved memory management with proper tensor disposal
• Enhanced Error Handling: More informative error messages and validation

Improvements

• 16KB Android Page Size Support: Compatible with Google Play's 16KB requirement for Android 15+ devices
• Simplified ONNX session initialization (no manual buffer loading required)
• Better documentation with inline migration comments
• Improved code organization and separation of concerns
• Added comprehensive architecture documentation in lib/src/README.md

Important Notes

• Isolate Support: This package cannot be used with Dart isolates because ONNX model loading requires Flutter asset access. See WHY_NO_ISOLATES.md for detailed explanation.
• Async/Await: Use async processing on the main isolate - it provides non-blocking behavior without isolate complexity
• Migration Guide: All changes are marked with // Migration: comments in the code

API Changes

• Session management simplified (internal changes, no user-facing API changes)
• Tensor creation and disposal updated to new API (handled internally)

Files Added

• lib/src/services/onnx_session_manager.dart - Session lifecycle management
• lib/src/utils/image_processor.dart - Image processing utilities
• lib/src/utils/mask_processor.dart - Mask manipulation utilities
• lib/src/utils/background_composer.dart - Background composition
• lib/src/README.md - Architecture documentation
• WHY_NO_ISOLATES.md - Isolate limitations explanation

Migration from v1.x

If upgrading from v1.x:

1. No code changes required! The dispose call remains the same:

   @override
   void dispose() {
     BackgroundRemover.instance.dispose();
     super.dispose();
   }
   

   Note: Even though dispose() is async internally, you should not await it in your widget's dispose method because Flutter's dispose must be synchronous.

2. The public API remains completely backward compatible!

1.0.0

Fix

• Minor Bug fix
• Readme file updated

Fix

• Minor Bug fix
• Readme file updated

Change log

• Implemented threshold-based segmentation to refine background removal.
• Integrated a smooth edge method using bilinear interpolation and average neighboring to improve output quality and reduce harsh edges.
• Implemented edge enhancement for mask refinement using a Sobel-like gradient detection method.

Added

• Implemented threshold-based segmentation to refine background removal.
• Integrated a smooth edge method using bilinear interpolation and average neighboring to improve output quality and reduce harsh edges.
• Implemented edge enhancement for mask refinement using a Sobel-like gradient detection method.

What's Changed

• Added addBackground function to add custom color background
• Modified Readme file

Feat

• Added addBackground function to change background of color

Full Changelog: https://github.com/Netesh5/image_background_remover/commits/0.0.5

Fix

• Minor bug fixes

Fix

• Solved ONNX session creation error

Added

• Added assets file.

Fix

• Removed incompatible platform support.

Added

• Initial release of the Background Remover Service Flutter package.
• ONNX Runtime integration for background removal using the onnx model.
• Functions to initialize ONNX session:
  • initializeOrt(): Initialize the ONNX environment and session.
• Image processing capabilities:
  • removeBg(Uint8List imageBytes): Removes the background from an input image and returns the image with a transparent background.
  • _resizeImage(): Resizes an image to 320x320 for ONNXy model compatibility.
  • _imageToFloatTensor(): Converts RGBA image data into a normalized float tensor for model input.
  • _applyMaskToOriginalSizeImage(): Applies the generated mask back to the original image size.
• Utility methods:
  • resizeMask(): Resizes the ONNX output mask to match the original image dimensions.
• Designed for cross-platform support (iOS, Android, Web, and Desktop).