# deepstream-yolov5

**Repository Path**: FIRC/deepstream-yolov5

## Basic Information

- **Project Name**: deepstream-yolov5
- **Description**: an implementation of yolov5 running on deepstream5
- **Primary Language**: C++
- **License**: MIT
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 1
- **Created**: 2022-01-10
- **Last Updated**: 2022-04-15

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

## DeepStream YOLOv5

An implementation of YOLOv5 running on DeepStream 5

## Requirements

```
CUDA 10.2
TensorRT 7.1
DeepStream 5
```
### or 
```
CUDA 11.1
TensorRT 7.2.1.6
DeepStream 5.1
```

### How to Use

Enter $ROOT of this repository.

* Step1: Prepare the wts file of YOLOv5s model follow instructions [tensorrtx/yolov5](https://github.com/wang-xinyu/tensorrtx/tree/master/yolov5), then put the wts file into $ROOT/data folder and rename as `yolov5s.wts`.
* Step2: Enter $ROOT/source folder, modify `EXFLAGS` and `EXLIBS` in `Makefile` corresponding to your installed TensorRT library path, run `make` command to compile the run-time library.
* Step3: Back to $ROOT folder, run `deepstream-app -c configs/deepstream_app_config_yolov5s.txt` command.
* Step4: rename the generated engine file `model_b1_gpu0_fp16.engine` as `yolov5s_b1_gpu0_fp16.engine` for reuse.

## Acknowledgements

* [https://github.com/wang-xinyu/tensorrtx](https://github.com/wang-xinyu/tensorrtx)