Model Card for traffic-accident-detection-yolo11x

This model is a fine-tuned version of the YOLOv11x object detection model, specialized for traffic accident detection using real-world surveillance data. It can detect and localize two main object categories: accident and vehicle in images or video frames.

Model Details

Model Description

• Developed by: Uppada Enos
• Shared by: Uppada Enos
• Model type: Object Detection (YOLOv11x-based)
• License: MIT
• Finetuned from model: Ultralytics/YOLO11

Uses

Direct Use

This model can be directly used for:

• Real-time traffic surveillance
• Automatic incident detection in smart cities
• Analyzing video streams for safety monitoring

Out-of-Scope Use

• Not designed for multi-class general object detection
• Not validated for use in critical autonomous driving systems

How to Get Started with the Model

from ultralytics import YOLO

# Load the fine-tuned model
model = YOLO('path/to/best.pt')

# Run inference on an image
results = model('path/to/image.jpg')

# Visualize the results
results.show()

Training Details

Training Data

Dataset used: Roboflow - Traffic Accident Detection

Training Hyperparameters

• Epochs: 61
• Batch size: 16
• Learning rate: 0.001
• Optimizer: SGD
• Image size: 640
• Augmentations: Mosaic, mixup, HSV shift, scaling, translation

Evaluation

Metrics

Model Card Authors

Uppada Enos, with fine-tuning using Ultralytics' YOLOv11x architecture.

Model Card Contact

For issues or contributions, reach out via the Hugging Face discussion tab or contact: enosuppada2005@gmail.com