Evaluation of Resource-Efficient Crater Detectors on Embedded Systems

TL;DR

The paper tackles onboard Mars crater detection using YOLO networks on resource-constrained embedded hardware. It trains and benchmarks YOLOv5 and YOLOv8 on a Mars crater dataset, evaluating across CPUs, GPUs, TPUs, and FPGAs with power and latency metrics. Key findings show favorable power efficiency on 5–15 W devices like Edge TPU and Orin, while Versal offers higher throughput at the cost of accuracy after deployment modifications. The work provides deployment pipelines and an open codebase, demonstrating the feasibility of autonomous navigation and geological exploration using real-time crater detection on spaceborne edge devices.

Abstract

Real-time analysis of Martian craters is crucial for mission-critical operations, including safe landings and geological exploration. This work leverages the latest breakthroughs for on-the-edge crater detection aboard spacecraft. We rigorously benchmark several YOLO networks using a Mars craters dataset, analyzing their performance on embedded systems with a focus on optimization for low-power devices. We optimize this process for a new wave of cost-effective, commercial-off-the-shelf-based smaller satellites. Implementations on diverse platforms, including Google Coral Edge TPU, AMD Versal SoC VCK190, Nvidia Jetson Nano and Jetson AGX Orin, undergo a detailed trade-off analysis. Our findings identify optimal network-device pairings, enhancing the feasibility of crater detection on resource-constrained hardware and setting a new precedent for efficient and resilient extraterrestrial imaging. Code at: https://github.com/billpsomas/mars_crater_detection.

Figures (2)

• Figure 1: Crater Area Distribution from Mars dataset aichallenge
• Figure 2: Crater detection AP$_{50}$ on Mars dataset aichallenge for distinct YOLOv8 versions across different architectures