Human Release v2.5

New

• New frame change detection algorithm used for cache determination
  based on temporal input difference
• New built-in Tensorflow profiling human.profile
• New optional input histogram equalization config.filter.equalization
  auto-level input for optimal brightness/contrast
• New event-baseed interface human.events
• New configuration validation human.validate
• New input compare function human.compare
  this function is internally used by human to determine frame changes and cache validation
• New custom built TFJS for browser bundled version
  result is a pure module with reduced bundle size and include built-in support for all backends
  note: nobundle and node versions link to standard @tensorflow packages

Changed

• Default configuration values have been tuned for precision and performance
• Supports all built-in modules on all backends
  via custom implementation of missing kernel ops
• Added additional hand gestures
• Performance and precision improvements
  • face, hand
  • gestures modules
  • face matching
• Fix ReactJS compatibility
• Better precision using WASM
  Previous issues due to math low-precision in WASM implementation
• Full TS type definitions for all modules and imports
• Focus on simplified demo
  https://vladmandic.github.io/human/demo/typescript/

Human Release v2.4

New:

• Time based caching
  add skipTime in addition to skipFrames
• Utility class human.env
• Add optional Anti-Spoof module

Changed:

• Enhanced MoveNet post-processing
• Increase human.similarity resolution
• Remove old HandDetect and PoseNet models from default installation
• Multiple bug fixes
• Updated configuration default values

Internals:

• Update to TFJS 3.10.0
• Update to Node v17 for dev environment
• Updated TS type definitions
• Updated JSDoc comments and TypeDoc documentation
• Refactor ImageFX module
• Experimental human.custom.esm module with custom TFJS build

Human Release v2.3

Major release of Human:

• Default hand models is now new HandTrack (old configuration is supported)
• Support for BlazePose v2 models (old v1 models are no longer supported)
• New face matching structure and algorithms: human.match() and human.similarity() (breaking change)
• Workaround for GPU memory leak on Chrome 93/94/95 (other Chrome versions are not affected)
• Support for running Human in Docker containers
• New examples gallery
• Major code refactoring for face and body modules
• Performance and precision improvements for hand and body modules
• Experimental support for WebGPU
• Autodetect desired number of targets for body and hand based on detected number of faces
• Optimized interpolation method human.next()

Rereleased due to issues with 2.3.1 build

Human Release v2.2

release 2.2 with major code refactoring, multiple bug fixes and precision and performance improvements

Human Release v2.0

2.0.1

add release notes

Human Release v1.8

1.8.1

release 1.8 with major changes and tfjs 3.6.0

Human Release v1.4

1.4.1

1.4.1

Human Release v1

1.0.1

update

Human: Release v0.3

update

Human: Release v0.1

update