Streamlit Video Analyzer

This repository hosts a real-time video streaming and analysis app using Streamlit and YOLO. The app provides options for live video streams from various sources, and integrates YOLO for real-time object detection, pose estimation, and segmentation.

Table of Contents

• Installation
• Usage
• Directory Structure
• Scripts Overview

Installation

1. Clone the repository:

git clone https://github.com/username/Streamlit-Video-Analyzer.git cd Streamlit-Video-Analyzer

2. Install the required Python packages:

```bash
pip install -r requirements.txt

Usage

To start the app, ensure that you have the correct model files and video sources. You can run the app as follows:

Running the Pipeline

Execute the following command to start the pipeline:

streamlit run main.py

Arguments

• source: Select a video source from webcam, video files, or RTSP stream.
• model_type : Choose the model type for video analysis, including object detection, pose estimation, or segmentation.

Directory Structure

The project’s directory structure is as follows:

Streamlit-Video-Analyzer/
│
├── camera/
│   ├── capture_video.py
│   ├── capture_video_d.py
│   └── ...
├── models/
│   ├── yolov8.pt
│   ├── pose_estimationv8.pt
│   ├── Segmentationv8.pt
│
├── class/
│   ├── labels.txt
│   └── ...
├── main.py
├── read_json.py
├── prompt.json
├── requirements.txt
└── README.md

Scripts Overview

• **main.py: The main script that starts the Streamlit app and loads the interface.
• **capture_video.py: Contains the function to capture video from standard sources.
• **capture_video_d.py: Contains the function to capture video from a specific camera type (e.g., Depth Camera).
• **draw.py: draws the inferencing.
• **read_json.py: reads in json and over writes two locations
• **prompt.json: is the prompt
• **`models/': Directory for storing pre-trained YOLO models.
• **`classes/': Directory for storing text files of classes.

Scripts Overview

• **Video Source: Choose between webcam, RTSP stream, and video files.
• **image: saves image to a direstor
• **Model Type: Select the type of analysis model (Object Detection, Pose Estimation, or Segmentation).
• **Resolution: Customize resolution for the video stream.
• **Confidence Threshold: Set a confidence threshold for the model predictions.
• **Classes: Select the class text file