A Real-Time FPGA Accelerator Based on Winograd Algorithm for Underwater Object Detection

Abstract

Real-time object detection is a challenging but crucial task for autonomous underwater vehicles because of the complex underwater imaging environment. Resulted by suspended particles scattering and wavelength-dependent light attenuation, underwater images are always hazy and color-distorted. To overcome the difficulties caused by these problems to underwater object detection, an end-to-end CNN network combined U-Net and MobileNetV3-SSDLite is proposed. Furthermore, the FPGA implementation of various convolution in the proposed network is optimized based on the Winograd algorithm. An efficient upsampling engine is presented, and the FPGA implementation of squeeze-and-excitation module in MobileNetV3 is optimized. The accelerator is implemented on a Zynq XC7Z045 device running at 150 MHz and achieves 23.68 frames per second (fps) and 33.14 fps when using MobileNetV3-Large and MobileNetV3-Small as the feature extractor. Compared to CPU, our accelerator achieves 7.5×–8.7× speedup and 52×–60× energy efficiency.

Marine Object Detection using YOLOv4 Adapted Convolutional Neural Network

“…Huang et al [16] contributed to the field with a novel approach using a restructured single-shot multi-box detector for underwater object detection, offering insights into improving detection accuracy and efficiency. Cai et al [17] introduced a real-time FPGA accelerator based on the Winograd algorithm, emphasizing the importance of hardware acceleration for efficient underwater object detection. Haq & Saqlain [18] placed a critical emphasis on precision, while Zulqarnain & Saqlain [19] identified text readability.…”