pytorch_lite 4.2.4

Flutter package to help run pytorch lite models classification and yolov5 and yolov8

4.2.4

• fixed ios possible memory leak (by cyrillkuettel)

4.2.3

• fixed export of enums and classes

4.2.2

• upgrading pytorch android from 1.13.1 to 2.1.0
• ios is still LibTorch 1.13.0.1 since its the last one

4.2.1

• fix classification bug on android on native preprocessing
• Making model output dynamically calculated (yolov5,yolov8) to allow input sizes other than 640X640

4.2.0+2

• Fix formatting

4.2.0+1

• Adding license
• Adding docstring for all functions

4.2.0

• Fixing ios camera image decoding
• Converting package to use native methods instead of ffi (fixing ios)
• Adding parameters to allow choosing between imageLib or native preprocessing
• Adding better integration testing

4.0.0

• image processing is done using opencv to improve performance
• camera image decoding is done using opencv to improve performance
• camera images can be ran using getCameraImage* methods
• improving the performance of object detection by moving all the processing to isolate
• Fixed: CAMERA image boxes are shifted upwards in example
• Improving example

3.0.4

• improved speed of inference
• improve build speed on android

3.0.3

• windows extracting using tar instead of powershell since powershell is so slow

3.0.2

• improving cmake for windows extraction (thanks https://github.com/ZhaoXinZhang for pointing it out).

3.0.1

• fix android build on Windows.

3.0.0+1

• Converted to work on ffi (for improved performance, and ios support)
• Improved isolates to stop ui frame drops
• Added ImageUtilsIsolate to process camera images in isolate instead of ui thread
• Updated Camera example
• Better memory usage and freeing
• Fixed camera example usage
• Breaking changes
• getImagePredictionFromBytesList Removed, (check camera example for new usage)

2.0.5

• fixed dart analyses problems (renamed some variables to follow convention)

2.0.4

• removed unnecessary prints

2.0.3

• fixed yolov8 bad performance

2.0.2

• com.facebook.soloader:nativeloader from 0.8.0 to 0.10.5

2.0.1

• upgraded to pigeon 9.0.0
• adding yolov8 support, thanks to https://github.com/atanasko/android-demo-app
• updated camera,image to latest version in example
• converted to using pytorch_android version to 1.12

2.0.0

• breaking change: Rect is now PyTorchRect to avoid conflicts in ios (when it is implemented)
• upgraded to pigeon 4.2.0

1.1.2

• Updated pytorch_android_lite version to 1.12.2

1.1.1

• Updated pytorch_android_lite version to 1.11

1.1.0

• Added get prediction from bytes list to run on camera image
• Added example to camera image prediction (thanks to KingWu)

1.0.6

• Updated dependencies
• Fixed release not working and Added to readme the solution

1.0.5

• Made all functions take image bytes instead of image file (to avoid storing stream in storage)

1.0.4

• Fixed a bug (probabilities were wrong)

1.0.3

• Added getting classification as probabilities

1.0.2

• Initial release

1.0.1

• Making code run in background to avoid frame drops

1.0.0

• Made some optimizations

0.0.4

• implemented object detection native code

0.0.3

• Upgraded dependencies and made some optimizations

0.0.2

• Used base code from https://pub.dev/packages/pytorch_mobile/changelog

0.0.1

• Setup pigeon messages