AI-Vision

Overview

This project leverages OpenCV and a pre-trained deep learning model to recognize and classify objects in images, videos, and live webcam feeds. The model used in this project is based on SSD MobileNet V3, which is trained on the COCO dataset—a large-scale dataset with 80 common object classes. The project allows for object detection and displays the detected objects with bounding boxes and labels.

Features

• Image-based Object Detection: Detect objects in static images.
• Video-based Object Detection: Detect objects in video files.
• Live Webcam Object Detection: Detect objects using a connected webcam.
• Real-time Object Classification: Using a pre-trained model to classify detected objects in real time.

Prerequisites

Before running the project, make sure you have the following dependencies installed:

• Python 3.x
• OpenCV (for image and video processing)
• Matplotlib (for image visualization)
• A pre-trained model file (frozen_inference_graph.pb and ssd_mobilenet_v3_large_coco_2020_01_14.pbtxt) -> Both are provided above.

Install the required packages using pip:

pip install opencv-python

pip install matplotlib

pip install numpy

Files & Configuration

• (frozen_inference_graph.pb): The pre-trained model used for object detection.
• (ssd_mobilenet_v3_large_coco_2020_01_14.pbtxt): The configuration file for the model.
• (labels.txt): A text file containing the class labels (objects) that the model can detect, corresponding to the COCO dataset.
• Test images/videos: Example files like boy.jpg, stop_sign.jpg, london_street.mp4, or webcam input can be used for object detection.
• Choose the python file that you wish to run based on your input (i.e. picture, video or camera).

Enhanced Configuration

• Model Configuration: The model uses the MobileNet V3 SSD architecture, trained on the COCO dataset. You can modify the (confThreshold) to adjust the confidence level for object detection.
• Window Size: Adjust the window size where the video feed or image will be displayed by using (cv2.resizeWindow()).

Conclusion

This project demonstrates how artificial intelligence, specifically deep learning and computer vision, can be applied to detect and classify objects in real-time from images, videos, and webcam streams. You can customize the code further to add more advanced features like tracking, multi-object recognition, or integration with other AI systems.