human/TODO.md

57 lines
1.8 KiB
Markdown
Raw Normal View History

2021-03-03 18:04:59 +01:00
# To-Do list for Human library
2022-06-02 16:39:53 +02:00
## Work-in-Progress
2021-04-25 19:16:04 +02:00
2022-06-02 16:39:53 +02:00
N/A
2021-09-26 16:09:30 +02:00
2022-06-02 16:39:53 +02:00
<hr><br>
2022-06-02 16:39:53 +02:00
## Exploring
2022-09-27 17:51:55 +02:00
- **Optical flow** for intelligent temporal interpolation
2022-06-02 16:39:53 +02:00
<https://docs.opencv.org/3.3.1/db/d7f/tutorial_js_lucas_kanade.html>
2022-09-27 17:51:55 +02:00
- **CLAHE** advanced histogram equalization for optimization of badly lit scenes
- **TFLite** models
2022-06-02 16:39:53 +02:00
<https://js.tensorflow.org/api_tflite/0.0.1-alpha.4/>
2022-09-27 17:51:55 +02:00
- **Principal Components Analysis** for reduction of descriptor complexity
<https://github.com/vladmandic/human-match/blob/main/src/pca.js>
- **Temporal guidance** for face/body segmentation
2022-06-02 16:39:53 +02:00
<https://github.com/PeterL1n/RobustVideoMatting>
<hr><br>
2021-07-31 13:43:50 +02:00
2021-08-12 15:31:16 +02:00
## Known Issues
2022-07-25 14:33:07 +02:00
2022-06-02 16:39:53 +02:00
### Face with Attention
2021-07-31 13:43:50 +02:00
2022-08-30 16:28:33 +02:00
`FaceMesh-Attention` is not supported in browser using `WASM` backend due to missing kernel op in **TFJS**
### Object Detection
`NanoDet` model is not supported in in browser using `WASM` backend due to missing kernel op in **TFJS**
2021-11-16 19:07:44 +01:00
2022-06-02 16:39:53 +02:00
### WebGPU
Experimental support only until support is officially added in Chromium
Enable via <chrome://flags/#enable-unsafe-webgpu>
2021-07-31 13:43:50 +02:00
2022-06-02 16:39:53 +02:00
### Firefox
Running in **web workers** requires `OffscreenCanvas` which is still disabled by default in **Firefox**
Enable via `about:config` -> `gfx.offscreencanvas.enabled`
<hr><br>
2022-06-02 16:39:53 +02:00
## Pending Release Changes
2022-07-18 14:22:19 +02:00
2022-09-17 23:19:51 +02:00
- New API [`human.video()`](https://vladmandic.github.io/human/typedoc/classes/Human.html#video)
Runs continous detection of an input video instead of processing each frame manually
2022-09-21 21:31:17 +02:00
- New simple demo [*Live*](https://vladmandic.github.io/human/demo/video/index.html)
- New advanced demo using BabylonJS <https://vladmandic.github.io/human-bjs-vrm>
2022-09-17 23:19:51 +02:00
- Enable model cache when using web workers
2022-09-21 21:31:17 +02:00
- Fix for `face.rotation` interpolation
2022-09-17 23:19:51 +02:00
- Improve NodeJS resolver when using ESM
2022-09-25 16:15:47 +02:00
- Update demo `demo/faceid`
2022-09-19 16:46:11 +02:00
- Update demo `demo/nodejs/process-folder.js`
2022-09-21 21:31:17 +02:00
and re-process `/samples`