Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Rapid Deployment of Domain-specific Hyperspectral Image Processors with Application to Autonomous Driving

Jon Gutiérrez-Zaballa, Koldo Basterretxea, Javier Echanobe, Óscar Mata-Carballeira, M. Victoria Martínez

TL;DR

The paper describes in detail the steps followed to redesign and customize a successfully trained HSI segmentation lightweight FCN that was previously tested on a high-end heterogeneous multiprocessing system-on-chip (MPSoC) to accommodate it to the constraints imposed by a low-cost SOM.

Abstract

The article discusses the use of low cost System-On-Module (SOM) platforms for the implementation of efficient hyperspectral imaging (HSI) processors for application in autonomous driving. The work addresses the challenges of shaping and deploying multiple layer fully convolutional networks (FCN) for low-latency, on-board image semantic segmentation using resource- and power-constrained processing devices. The paper describes in detail the steps followed to redesign and customize a successfully trained HSI segmentation lightweight FCN that was previously tested on a high-end heterogeneous multiprocessing system-on-chip (MPSoC) to accommodate it to the constraints imposed by a low-cost SOM. This SOM features a lower-end but much cheaper MPSoC suitable for the deployment of automatic driving systems (ADS). In particular the article reports the data- and hardware-specific quantization techniques utilized to fit the FCN into a commercial fixed-point programmable AI coprocessor IP, and proposes a full customized post-training quantization scheme to reduce computation and storage costs without compromising segmentation accuracy.

Rapid Deployment of Domain-specific Hyperspectral Image Processors with Application to Autonomous Driving

TL;DR

The paper describes in detail the steps followed to redesign and customize a successfully trained HSI segmentation lightweight FCN that was previously tested on a high-end heterogeneous multiprocessing system-on-chip (MPSoC) to accommodate it to the constraints imposed by a low-cost SOM.

Abstract

The article discusses the use of low cost System-On-Module (SOM) platforms for the implementation of efficient hyperspectral imaging (HSI) processors for application in autonomous driving. The work addresses the challenges of shaping and deploying multiple layer fully convolutional networks (FCN) for low-latency, on-board image semantic segmentation using resource- and power-constrained processing devices. The paper describes in detail the steps followed to redesign and customize a successfully trained HSI segmentation lightweight FCN that was previously tested on a high-end heterogeneous multiprocessing system-on-chip (MPSoC) to accommodate it to the constraints imposed by a low-cost SOM. This SOM features a lower-end but much cheaper MPSoC suitable for the deployment of automatic driving systems (ADS). In particular the article reports the data- and hardware-specific quantization techniques utilized to fit the FCN into a commercial fixed-point programmable AI coprocessor IP, and proposes a full customized post-training quantization scheme to reduce computation and storage costs without compromising segmentation accuracy.

Paper Structure

This paper contains 7 sections, 2 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. FCN model development
  3. Model quantization
  4. Network deployment and testing
  5. Processor implementation
  6. Experimental results and system validation
  7. Conclusions

Figures (5)

  • Figure 1: Input data distribution for the 25 spectral channels.
  • Figure 2: Packing two INT8 MAC with one DSP48E2 slice (adapted from fu2016deep).
  • Figure 3: DDR data ports read/write rates as the DPU segments 20 images.
  • Figure 4: Segmentation of the lightweight FCN on interurban scenarios. Rows: top, ground-truth; center, segmentation and bottom, false color. Columns: far left, image 617 (winter, cloudy, morning); left, image 209 (spring, rainy, morning); right, image 665 (winter, cloudy, afternoon) and far right, image 633 (autumn, cloudy, midday).
  • Figure 5: Segmentation of the lightweight FCN on highway and urban scenarios. Rows: top, ground-truth; center, segmentation and bottom, false color. Columns: far left, image 637 (winter, sunny, midday, urban); left, image 626 (winter, cloudy, morning, urban); right, image 18 (summer, sunny, midday, highway) and far right, image 307 (winter, cloudy, midday, highway).