projects/human_detection/ReadMe.md

# HUMAN DETECTION DEMO

### 1. Prepare data and weights:
The weights and sample test video files can be downloaded from:
+ NAS at `/5_project_internal/19_Demo_App/human_detection`.
+ On HPC2,`/data/cc-demo/human_detection/weights/`

If you run the demo on other machine, please download the files and put it the same path as HPC2.
From now on, we will assume that you have data and weights in the following structure:
```
/data/cc-demo/human_detection/
├── sample_inputs
│   ├── 1.mp4
│   ├── 2.mp4
│   └── 3.mp4
└── weights
        └── mmyolov8_s_human_dynamic_shape.onnx
            mmyolov8_s_human_DBsx640x800.trt


```
### 2. Compile ONNX & TRT model 
#### 2.1. Start docker container
For Developer that want to add more functions to the codebase, and use Docker as Developing environment, use:
```
bash docker_run.sh 
docker attach <docker_container_name>
```
After attaching into docker, do the following step to export onnx and trt model
```
bash projects/human_detection/export_onnx_trt/export_trt_mmyolov8.sh
```
It will take about 20minutes to comple ONNX & TRT model, and the file will be stored at folder `/data/human_detection\deploy`

### 3. Start Gradio 
a. From inside Docker
```
bash projects/human_detection/demo_app.sh
```
b. From Host machine
```
bash projects/human_detection/docker_run.sh
```