Reproduce “SpiderCam: Low-Power Snapshot Depth from Differential Defocus” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Marcos A. Ferreira, Tianao Li, John Mamish, Josiah Hester, Yaman Sangar, Qi Guo, Emma Alexander
SpiderCam presents an FPGA-based snapshot depth-from-defocus camera that generates sparse depth maps in real-time, achieving a total system power consumption of 624 mW at 32.5 FPS. This system represents the first passive FPGA-based 3D camera to operate at sub-Watt total power, demonstrating a working range of 0.45 m to 0.97 m with less than 10% mean absolute error.
Key insights
- Differential Depth from Defocus (DfDD) algorithms, when meticulously optimized for operation count and memory locality, can deliver real-time depth sensing within sub-Watt power constraints on FPGAs.
- Mitigating optical aberrations inherent in compact, low-cost optics requires adaptive calibration, where parameters are allowed to vary spatially across the image, significantly extending the system's practical working range.
- Hardware implementations must adopt memory-local streaming and carefully tailored arithmetic (e.g., hybrid fixed/floating-point, integer-coefficient kernels) to fit complex algorithms onto low-power FPGAs without requiring extensive memory buffering.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.