Precision, speed, and efficiency in detecting vehicles across diverse environments.
Welcome to Vehicle Object Detection using YOLOv8n! This project leverages the YOLOv8n model to perform high-accuracy vehicle detection. Utilizing the JUIVCDv1 dataset, this project addresses real-world challenges in vehicle classification, achieving exceptional results with minimal computational overhead.
"Simplifying vehicle detection, enhancing every journey."
- 🚗 Real-Time Vehicle Detection: Utilizing the state-of-the-art YOLOv8n architecture.
- 🧠 Robust Model: Achieves 97.4% mAP on the test set, ensuring high precision.
- 🔄 Comprehensive Data Processing: Addressing dataset inconsistencies for better accuracy.
Vehicle Object Detection using YOLOv8n involves several critical steps from data preparation to model evaluation:
- YOLOv8n (Ultralytics)
- Google Colab for training
- Roboflow for data annotation and pre-processing
- Python Libraries: OpenCV, NumPy, Pandas
-
Data Preparation:
- Initially the JUIVCDv1 dataset from downloaded from Kaggle, consisting of images and annotation.
- It has many data issues like annotation mismatches, image format inconsistencies, and incorrect class mappings.
- So it is exported to roboflow for reannotation,resolved issues and then exported data in yolo format.
- The data folder has preprocessed data imported from roboflow.
-
Training Configuration:
- Batch Size: 16
- Epochs: 50
- Image Size: 640x640
- Optimizer: Auto
-
Model Evaluation:
- Achieved 97.4% mAP50, 87.6% mAP50-95, 96.6% precision and 96% recall on test data.
- mAP50: 97.4%
- mAP50-95: 87.6%
- Precision: 96.6%
- Recall: 96%
These metrics highlight the model's robustness in detecting and classifying vehicles with high confidence.
Example of YOLOv8n accurately detecting vehicles with high confidence.
