Enhancing Object Detection Framework (MMDetection/YOLOv3)
Budget: £250 – £750 GBP
Dual-Branch Scene Text Detection Model Development (MMDetection/YOLOv3）
Project goal
To improve the existing object detection framework by integrating a dual-branch architecture and achieving a target F1 score of at least 0.85.
Scope of work
Implement dual-branch architecture with RGB and frequency-domain branches within the existing MMDetection/YOLOv3 framework.
Develop a fusion module for feature combination between branches.
Train and evaluate the model on the Tampered_IC13 dataset (and potentially similar datasets) to meet the target F1 score ≥ 0.85.
Deliver:
Trained model weights (.pth)
Training and evaluation logs
Detailed evaluation metrics (Precision, Recall, F1)
Integration/usage guide to reproduce results
Optimize performance and conduct experiments to exceed baseline performance.
Required skills: PyTorch, MMDetection, YOLOv3, and frequency-domain image processing (e.g., FFT, DCT).
Developer expertise
Image processing, Anomaly detection
Programming language
Python
Project goal
To improve the existing object detection framework by integrating a dual-branch architecture and achieving a target F1 score of at least 0.85.
Scope of work
Implement dual-branch architecture with RGB and frequency-domain branches within the existing MMDetection/YOLOv3 framework.
Develop a fusion module for feature combination between branches.
Train and evaluate the model on the Tampered_IC13 dataset (and potentially similar datasets) to meet the target F1 score ≥ 0.85.
Deliver:
Trained model weights (.pth)
Training and evaluation logs
Detailed evaluation metrics (Precision, Recall, F1)
Integration/usage guide to reproduce results
Optimize performance and conduct experiments to exceed baseline performance.
Required skills: PyTorch, MMDetection, YOLOv3, and frequency-domain image processing (e.g., FFT, DCT).
Developer expertise
Image processing, Anomaly detection
Programming language
Python