Refresh release:

• Toolkit upgrade: TypeScript 5.3.3, TSLib 2.6.2, TFJS 4.17.0, ESLint 8.55.0, ESBuild 0.19.22
• Custom CI platform major refresh to make it compatible with new toolkits
• Add new public config variables, face.detector.scale and face.iris.scale

Refresh Release

• major toolkit upgrade:
  node v20, npm v10, typescript v5.1, tfjs: 4.11
• add electron environment detection
• fix gender-ssrnet-imdb
• fix movenet-multipose
• add optional crop values for emotion, description, ssrnet and gear models

Optimizations:

• Enabled high-resolution optimizations
  Internal limits are increased from 2k to 4k
• Enhanced device capabilities detection
  See human.env.[agent, wasm, webgl, webgpu] for details
• If config.backend is not set, Human will auto-select best backend
  based on device capabilities
• Enhanced support for webgpu
• Reduce build dependencies
  Human is now 30% smaller :)
  As usual, Human has zero runtime dependencies,
  all devDependencies are only to rebuild Human itself
• Default hand skeleton model changed from handlandmark-full to handlandmark-lite
  Both models are still supported, this reduces default size and increases performance

Features:

• Add draw label templates
  Allows easy customization of results labels drawn on canvas
• Add config.filter.autoBrightness (enabled by default)
  Per-frame video on-the-fly brightness adjustments
  Which significantly increases performance and precision in poorly lit scenes
• Add new demo face detect
• Improved config.filter.equalization (disabled by default)
  Image and video on-demand histogram equalization
• Support selecting specific video source when multiple cameras are present
  See human.webcam.enumerate()
• Updated algorithm to determine distance from camera based on iris size
  See human.result.face[n].distance

Architecture:

• Upgrade to TFJS 4.1 with strong typing
  see notes on how to use
• TypeDef refactoring
• Re-architect human.models namespace for better dynamic model handling
  Added additional methods load, list, loaded, reset
• Repack external typedefs
  Removes all external typedef dependencies
• Refactor namespace exports
  Better TypeDoc specs
• Add named export for improved bundler support when using non-default imports
• Cleanup Git history for dist/typedef/types
• Cleanup @vladmandic/human-models
• Support for NodeJS v19
• Upgrade to TypeScript 4.9
• Support for dynamic module load in NodeJS
  See https://vladmandic.github.io/human/demo/nodejs/node-bench

Breaking changes:

• Replaced result.face[n].iris with result.face[n].distance
• Replaced human.getModelStats() with human.models.stats()
• Moved human.similarity, human.distance and human.match to namespace human.match.*
• Obsolete human.enhance()
• Obsolete human.gl
• Renamed model mb3-centernet to centernet

TL;DR: Several new methods and models, but only breaking change is how human.segmentation is handled

• New methods human.webcam.*
  Enables built-in configuration and control of WebCam streams
• New method human.video()
  Runs continous detection of an input video
  instead of processing each frame manually using human.detect()
• New demo for webcam and video methods Live | Code
  Full HTML and JavaScript code in less than a screen
• Redesigned human.segmentation
  Breaking changes
• New model rvm for high-quality body segmentation in real-time
  Not part of default deployment, download from human-models
• New demo for segmentation methods Live | Code
  Full HTML and JavaScript code in less than a screen
• New advanced demo using BabylonJS and VRM Live | Code
• Update TypeDoc generation Link
• Update TypeDefs bundle generation Link
  No external dependencies
• Fix model caching when using web workers
• Fix face.rotation when using interpolation
• Improve NodeJS resolver when using ESM
• Update demo demo/typescript
• Update demo demo/faceid
• Update demo demo/nodejs/process-folder.js
  and re-process /samples Link

Human v2.10 is flagged as major release purely due to number of changes and fixes, but there are no breaking changes

• Update TFJS to 3.20.0
• Update TypeScript to 4.8
• Switch default backend from custom humangl to webgl
  Stability and availability of features in standard TFJS allows to switch back
• Add InsightFace model as alternative for face embedding/descriptor detection
  Compatible with multiple variations of InsightFace models
  Configurable using config.face.insightface config section
  See demo/faceid/index.ts for usage
  Models can be downloaded from https://github.com/vladmandic/insightface
• Add human.check() which validates all kernel ops for currently loaded models with currently selected backend
  Example: console.error(human.check());
• Add config.softwareKernels config option which uses CPU implementation for missing ops
  Disabled by default
  If enabled, it is used by face and hand rotation correction (config.face.rotation and config.hand.rotation)
• Add underlying tensorflow library version detection when running in NodeJS to
  human.env and check if GPU is used for acceleration
  Example: console.log(human.env.tensorflow)
• Treat models that cannot be found & loaded as non-critical error
  Instead of creating runtime exception, human will now report that model could not be loaded
• Improve human.reset() method to reset all config values to defaults
• Host models in https://github.com/vladmandic/human-models
  Models can be directly used without downloading to local storage
  Example: modelBasePath: 'https://vladmandic.github.io/human-models/models/'
• Allow hosting models in Google Cloud Bucket
  Hosted models can be directly used without downloading to local storage
  Example: modelBasePath: 'https://storage.googleapis.com/human-models/'
• Stricter linting rules for both TypeScript and JavaScript
  See ./eslintrc.json for details
• Enhanced type safety across entire library
• Fix MobileFaceNet model as alternative for face embedding/descriptor detection
  Configurable using config.face.mobilefacenet config section
• Fix EfficientPose module as alternative body detection
• Fix NanoDet module as alternative object detection
• Fix demo/multithread/node-multiprocess.js demo
• Fix human.match when using mixed descriptor lengths
• Fix WASM feature detection issue in TFJS with Edge/Chromium
  Example: console.log(human.env.wasm)
• Reorganized init & load order for faster library startup
• Increased NodeJS test coverage
  Run using: npm run test
  Runs tests for tfjs-node, tfjs-node-gpu and wasm
• Increased Browser test coverage
  Run using: demo/browser.html
  Runs tests for webgl, humangl, webgpu and wasm
  Runs tests for ESM and IIFE versions of library
• Add config.flags settings to allow setting of custom TFJS flags during backend configuration
• Increase availability of alternative models
  See models/model.json for full list
• Update profiling methods in human.profile()
• Update project dependencies to latest versions

• Install production-only dependencies by default
  Results in a must faster and smaller human installation
  To install all dependencies use npm install @vladmandic/human --production=false
• Switch to production @tensorflow/tfjs for browsers
  tfjs has stabilized in recent versions so its not necessary to run a custom bundle anymore
• Add webview support
• Add getModelStats method
• Extract model stats in build process
• TypeDoc fixes
• Add face contours to results
• Improve face compare in demo app
• Update dependencies and tfjs
• gear model fixes

• Updated FaceMesh-Landmarks models
• Added FaceMesh-with-Attention model is disabled by defaults, enable using
  config.face.mesh.attention = true
• If FaceMesh-with-Attention model is enabled, Iris model gets disabled
  as its functionality is superseded by attention model
• Results include more detailed face mesh annotations
  result.face[].annotations
• Update benchmarks: demo/benchmark
• Slight performance improvement with canvas reads using WASM backend
• Default to RTM version of WASM binaries
• Enhanced WebGPU compatibility
• Support for NodeJS v18
• Updated Known Issues
  See https://github.com/vladmandic/human/blob/main/TODO.md

• Add MediaPipe FaceMesh-with-Attention model
  model is available in @vladmandic/human-models repository
  to enable, set config.face.attention = true
  model replaces iris, eyes and lips keypoints with high-detailed ones
• Add model pre-compile phase to warmup method
  result is speed-up to first inference by around ~30% for browser environments
• Changed default face crop from 120% to 140%
  to better utilize caching between frames and face mesh mapping
• Refactor draw methods into separate modules and fix coloring function
• Add highlights to attention keypoints
  enabled when both points and attention draw are enabled:
  human.draw.options.drawAttention = true and human.draw.options.drawPoints = true
• Add ElectronJS demo:
  see https://github.com/vladmandic/human-electron
• Enhanced 3D demos:
  see https://github.com/vladmandic/human-motion
• Support for 4K input
• Update TFJS to 3.16.0
• Update build platform and dependencies

• Add model cache hander using IndexDB in browser environments
  see config.cacheModels setting for details
• Add additional demos
  human-motion and human-avatar
• Updated samples image gallery
• Allow monkey-patching fetch in NodeJS
  If fetch function is defined, it can be used to load models
  from HTTP/HTTPS URLs regardless of tfjs platform support
• Fix BlazeFace detections when face is partially occluded
• Fix BlazeFace box scaling
• Fix HandTrack tracking when hand is in front of face
• Fix ElectronJS compatibility issues
• Fix body keypoints interpolation algorithm
• Updated BlazePose calculations
• Changes to BlazePose and HandLandmark annotations
• Strong typing for string enums
• Updated TFJS

New:

• Demo demos/faceid that utilizes multiple algorithms to validate input before triggering face recognition
• Type definitions rollup for Human and TFJS
• Optional module liveness
  checks if input appears to be a real-world live image or a recording
  best used together with module antispoofing that checks if input appears to have a realistic face
• Face masking option in face.config.detector.mask
  result is shading of face image outside of face area which is useful for increased sensitivity of other modules that rely on detected face as input
• Face crop option in face.config.detector.cropFactor
  result is user-definable fine-tuning for other modules that rely on detected face as input

Other:

• Update TFJS to 3.12.0
• Documentation overhaul
• Improved Safari compatibility
• Improved similarity and match score range normalization
• Improved error handling
• Improved VSCode out-of-the-box experience
• Fix for optional gear, ssrnet, mobilefacenet modules
• Fix for Firefox WebGPU compatibility issue
• Fix face detect box scale and rotation
• Fix body interpolation
• Updated blazepose implementation
• Strong typing for all string enums in config and results