Exploring Runtime Sparsification of YOLO Model Weights During Inference

Abstract

In the pursuit of real-time object detection with constrained computational resources, the optimization of neural network architectures is paramount. We introduce novel sparsity induction methods within the YOLOv4-Tiny framework to significantly improve computational efficiency while maintaining high accuracy in pedestrian detection. We present three sparsification approaches: Homogeneous, Progressive, and Layer-Adaptive, each methodically reducing the model’s complexity without compromising its detection capability. Additionally, we refine the model’s output with a memory-efficient sliding window approach and a Bounding Box Sorting Algorithm, ensuring precise Intersection over Union (IoU) calculations. Our results demonstrate a substantial reduction in computational load by zeroing out over 50% of the weights with only a minimal 6% loss in IoU and 0.6% loss in F1-Score.