# SignX

https://signerx.github.io/SignX/

This is a CSLR model. In the future, it will be further improved to become an open-world SLT model. 

Currently, it has been pluginized, that is, the root directory contains viser_backend.py and index.js. These two files will enable this repository to be recognized by the control panel as a plugin. 

Therefore, users should first clone the panel, configure the panel environment, and then place this repository in the plugin folder.

## Environment Setup

SignX inference uses two conda environments:

- `signx-slt`: PyTorch / SMKD stage (RGB feature extraction)
- `slt_tf1`: TensorFlow 1.15 / SLTUNET stage (Recognition based on feature space)


```bash
conda env create -f env/signx-slt.yml
conda activate signx-slt
pip install -r env/signx-slt.txt

conda env create -f env/slt_tf1.yml
conda activate slt_tf1
pip install -r env/slt_tf1.txt
```

Quick check:

```bash
conda activate signx-slt
python -V

conda activate slt_tf1
python -V
```

## Operation Video

<video controls width="100%">
  <source src="asset/SignX_operate.mp4" type="video/mp4">
  Your browser does not support the video tag.
</video>

<p><a href="asset/SignX_operate.mp4">Download the operation video</a></p>