CaptureCount - Object Detector on Raspberry Pi

Note

• There is an alternative script named capturecount.py in the repository, designed to work on Ubuntu 23.10 on Raspberry Pi 5.
• Despite the name, the script is capable of detecting a wide range of objects as specified in the YOLO coco.names file.

Overview

capturecount-pi.py is a Python script designed for real-time object detection on Raspberry Pi 5, using the YOLO (You Only Look Once) model developed by Joseph Redmon. This powerful tool can identify and track various objects as defined in the coco.names file from the YOLO model. It's optimized for Pi OS on Raspberry Pi 5.

Features

• Real-time object detection using YOLOv3.
• Saves full-frame images upon object detection, rather than cropped images of detected objects.
• Utilizes a pre-trained YOLO model for efficient and accurate detection.
• Keeps track of detected objects and their counts.
• Outputs results in various formats including images, CSV files, and logs.

Prerequisites

• Raspberry Pi 5 with Pi OS installed.
• Python environment with necessary libraries (cv2, pandas, numpy, subprocess, os).
• YOLOv3 model files (yolov3.weights, yolov3.cfg) and coco.names.

Installation

1. Clone the repository to your Raspberry Pi.
2. Ensure Python and the required libraries are installed.
3. Download YOLOv3 model files and coco.names, and place them in the ./yolo directory within the script's folder.
4. Run capturecount-pi.py.

Usage

• Execute the script: python capturecount-pi.py
• The script continuously captures images and performs object detection.
• Detected objects are framed with bounding boxes in real-time display.
• Full-frame images are saved in the output folder when an object is detected.
• Object count per category is saved in object_counts.csv and total_object_counts.csv.
• A text log of total counts is also maintained in total_counts_log.txt.

Acknowledgements

Credits to Joseph Redmon @pjreddie for developing the YOLO: Real-Time Object Detection model, which was crucial in the creation of this project.