copilot/OpenSeeFace
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Some outdated information was left in the README.

Browse source
This commit is contained in:
Emiliana 2020-05-01 11:56:31 +02:00 • committed by GitHub
parent c9c5193a55
commit f47e78d53e
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 5 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

10
README.md
View file

@ -10,7 +10,7 @@ Thanks to [@Virtual_Deat](https://twitter.com/Virtual_Deat) and [@ENiwatori](htt
[Unity sample video](https://twitter.com/emiliana_vt/status/1210622149314203648) | [Sample video](https://www.youtube.com/watch?v=AOPHiAp9DBE) | [Sample video](https://www.youtube.com/watch?v=-cBSuHGdBWQ) | [Model comparison](https://www.youtube.com/watch?v=yMcJszUy7FA)
These sample videos use older versions of the tracking models.
These sample videos use older versions of the tracking models. Tracking accuracy of the current models is better.
# Usage
@ -82,9 +82,9 @@ To save both the trained model and the captured training data, type in a filenam
Four pretrained models are included. Using the `--model` switch, it is possible to select them for tracking. The given fps values are for running the model on a single face video on a single CPU core. Lowering the frame rate would reduce CPU usage by a corresponding degree.
* Model **0**: This is a very fast, low accuracy model. (68fps)
* Model **1**: This is a slightly slower model with better accuracy based on MobileNetV3. (59fps)
* Model **2**: This is a slower model with good accuracy based on ShuffleNetV2. (50fps)
* Model **3** (default): This is the slowest and highest accuracy model. It is based on MobileNetV3. (44fps)
* Model **1**: This is a slightly slower model with better accuracy. (59fps)
* Model **2**: This is a slower model with good accuracy. (50fps)
* Model **3** (default): This is the slowest and highest accuracy model. (44fps)
FPS measurements are from running on one core of my CPU.
@ -156,7 +156,7 @@ The algorithm is inspired by:
The MobileNetV3 code was taken from [here](https://github.com/rwightman/gen-efficientnet-pytorch).
For all training after the first model, a modified version of [Adaptive Wing Loss](https://github.com/tankrant/Adaptive-Wing-Loss) was used.
For all training a modified version of [Adaptive Wing Loss](https://github.com/tankrant/Adaptive-Wing-Loss) was used.
* [Adaptive Wing Loss for Robust Face Alignment via Heatmap Regression](https://arxiv.org/abs/1904.07399) by Xinyao Wang, Liefeng Bo, Li Fuxin