VisionISP: Repurposing the Image Signal Processor for Computer Vision Applications
This addresses the problem of inefficient data transmission for computer vision applications, particularly in autonomous driving, though it is incremental as it adapts existing ISP concepts.
The authors tackled the mismatch between human-optimized image signal processors (ISPs) and computer vision needs by proposing VisionISP, a repurposed ISP that reduces bit-depth and resolution while preserving relevant information, boosting object detection performance in autonomous driving.
Traditional image signal processors (ISPs) are primarily designed and optimized to improve the image quality perceived by humans. However, optimal perceptual image quality does not always translate into optimal performance for computer vision applications. We propose a set of methods, which we collectively call VisionISP, to repurpose the ISP for machine consumption. VisionISP significantly reduces data transmission needs by reducing the bit-depth and resolution while preserving the relevant information. The blocks in VisionISP are simple, content-aware, and trainable. Experimental results show that VisionISP boosts the performance of a subsequent computer vision system trained to detect objects in an autonomous driving setting. The results demonstrate the potential and the practicality of VisionISP for computer vision applications.