YOLOv8s Object Detection with TensorFlow Lite and GStreamer Utilizing NNStreamer

This repository contains a GStreamer pipeline implementation for performing object detection using the YOLOv8s (float32) model with TensorFlow Lite. The pipeline utilizes the nnstreamer plugin for seamless integration and efficient inference.

Requirements

• GStreamer (>= 1.18)
• nnstreamer (>= 1.1.0)

Installation

1. GStreamer: Follow the installation instructions provided by the GStreamer website: GStreamer Installation Guide

2. Install YoloV8:

   Install YoloV8:

   $ pip install ultralytics

   Export to tflite model

   from ultralytics import YOLO
   
   # Load a model
   model = YOLO("yolov8s.pt") # load a pretrained model
   
   # Export the model
   model.export(format="tflite", imgsz=320) # export the model to tflite format
   '''
   

3. nnstreamer: Follow the installation instructions provided by the nnstreamer GitHub repository: nnstreamer Installation Guide

Usage

1. Clone this repository:

   $ git clone https://github.com/shubhamnagla/nnstreamer_yolo.git

2. Run the GStreamer pipeline:

   $ ./nnstreamer_yolov8_pipeline.py

Video Demo

Acknowledgements

This project is based on the following libraries and frameworks:

• GStreamer: https://gstreamer.freedesktop.org/
• nnstreamer: https://github.com/nnstreamer/nnstreamer