Overview:
This project aims to track and count cars using the Deep Sort algorithm integrated with the YOLOv8l model. The system detects and tracks vehicles in real time, allowing for accurate counting across multiple vehicle classes. It was trained on a dataset of 4680 images, including 10 vehicle classes, and achieved a mean average precision (mAP) of 0.764.
Key Features:
1. YOLOv8 for Vehicle Detection:
The system uses YOLOv8l for highly accurate and fast real-time detection of multiple vehicle types, ensuring robust identification in different traffic conditions.
2. Deep Sort Algorithm for Tracking:
By integrating the Deep Sort algorithm, the system efficiently tracks vehicles across multiple frames, assigning unique IDs to each vehicle, preventing double counting or loss of tracking during motion.
3. Diverse Dataset:
Trained on 4680 images across 10 different vehicle classes, the system is equipped to handle varied traffic scenarios, enhancing the detection of multiple vehicle types, including cars, trucks, and buses.
4. Performance:
The model achieved a mean average precision (mAP) of 0.764, along with precision of 0.757 and recall of 0.687, reflecting the system’s reliability in detecting and tracking vehicles accurately.
Category:
Tags:
Links
Portfolio 
Leave a comment