Fast Scene Understanding for Autonomous Driving

2017/08/08 by Davy Neven, Neven, Davy, Bert De Brabandere +7 · 9 citations
Computer Science · #Advanced Image and Video Retrieval Techniques #Advanced Neural Network Applications #Multimodal Machine Learning Applications #cs.CV #cs.RO

paper · pdf · doi:10.48550/arxiv.1708.02550

Published at "Deep Learning for Vehicle Perception", workshop at the IEEE Symposium on Intelligent Vehicles 2017

arxiv created 2017/08/08 · arxiv updated 2017/08/10

Abstract

Most approaches for instance-aware semantic labeling traditionally focus on accuracy. Other aspects like runtime and memory footprint are arguably as important for real-time applications such as autonomous driving. Motivated by this observation and inspired by recent works that tackle multiple tasks with a single integrated architecture, in this paper we present a real-time efficient implementation based on ENet that solves three autonomous driving related tasks at once: semantic scene segmentation, instance segmentation and monocular depth estimation. Our approach builds upon a branched ENet architecture with a shared encoder but different decoder branches for each of the three tasks. The presented method can run at 21 fps at a resolution of 1024x512 on the Cityscapes dataset without sacrificing accuracy compared to running each task separately.

Cited by

• Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation
• Learning to Fuse Things and Stuff
• RaidaR: A Rich Annotated Image Dataset of Rainy Street Scenes
• RFBNet: Deep Multimodal Networks with Residual Fusion Blocks for RGB-D Semantic Segmentation
• Monocular Outdoor Semantic Mapping with a Multi-task Network
• Advancing Complex Wide-Area Scene Understanding with Hierarchical Coresets Selection
• Spatial Sampling Network for Fast Scene Understanding
• A Survey of Foundation Models for IoT: Taxonomy and Criteria-Based Analysis
• AVROBUSTBENCH: Benchmarking the Robustness of Audio-Visual Recognition Models at Test-Time
• Multiple-Human Parsing in the Wild

Related

• DSNet for Real-Time Driving Scene Semantic Segmentation
• Real-time semantic segmentation on FPGAs for autonomous vehicles with hls4ml
• Deep Dual-resolution Networks for Real-time and Accurate Semantic Segmentation of Road Scenes
• Spatial-Assistant Encoder-Decoder Network for Real Time Semantic Segmentation
• Real time backbone for semantic segmentation