image-recognisation-using-python-opencv

This Python project demonstrates real-time video analysis using Google's Gemini AI model and OpenCV. It captures video from a webcam or a specified video file, sends frames to the Gemini model for description, and displays the generated text on the video feed in real-time.

Key Features:

• Flexible Video Input: Supports both webcam and video file sources.
• Efficient Processing: Achieves near real-time processing with configurable frame skipping.
• AI-Powered Analysis: Integrates with Gemini AI for image analysis and description generation.
• Dynamic Text Overlay: Displays Gemini's responses directly on the video frame.
• Text Handling: Wraps text to fit the screen and manages display duration.
• Concise Codebase: Utilizes short, readable variables and minimal comments for clarity.
• Error Management: Provides basic error handling for common issues.

How It Works:

1. Capture: The application grabs video frames from a webcam or video file.
2. Process: Frames are sampled at a configurable rate (skip) and encoded to base64 for the Gemini API.
3. Analyze: The base64 image and a descriptive prompt are sent to the Gemini model.
4. Display: Gemini's text response is overlaid onto the video, with line wrapping.
5. Control: The text is displayed for a set duration before disappearing.
6. Real-Time Output: Processed frames are displayed immediately.

Usage:

1. Set up a virtual environment (recommended):

   python -m venv venv
   source venv/bin/activate  # On macOS/Linux
   venv\Scripts\activate   # On Windows

2. Install dependencies:

   pip install opencv-python google-generativeai

3. Obtain an API Key: Get your Google Gemini API key from Google AI Studio.
4. Configure API Key: Set the API_KEY variable in the script to your obtained key.
5. Run the script:

   python your_script_name.py  # Or python3 your_script_name.py

6. Configuration:
   • Adjust skip: The number of frames to skip between processing (e.g., skip=5 will process every 5th frame).
   • Adjust max_f: The maximum number of frames to process (useful for testing).
   • Adjust txt_dur: The duration in seconds the generated text is shown. These parameters can be modified in the proc_vid function call at the end of the script.
7. Change the Model: You can change the Gemini model using genai.GenerativeModel(<model_name>) in the script. The default model is gemini-2.0-flash-exp.
8. Use Video Files: To analyze a video file, set src='file' and specify the vid_path (e.g., vid_path='path/to/your/video.mp4') when calling the proc_vid function.
9. Exit: Press the 'q' key while the window has focus to close the window and stop the script.

Code Structure:

• your_script_name.py: The main Python script containing all the video processing and Gemini integration logic.

Important Notes:

• This project uses the gemini-2.0-flash-exp model by default. You can modify the code to use a different model by changing the genai.GenerativeModel() call.
• Remember to replace "YOUR_API_KEY" with your actual API key from Google AI Studio.
• This project requires a valid Google AI API key to access the Google Gemini services.