FasterSeg: Searching for Faster Real-time Semantic Segmentation

Abstract

We present FasterSeg, an automatically designed semantic segmentation network with not only state-of-the-art performance but also faster speed than current methods. Utilizing neural architecture search (NAS), FasterSeg is discovered from a novel and broader search space integrating multi-resolution branches, that has been recently found to be vital in manually designed segmentation models. To better calibrate the balance between the goals of high accuracy and low latency, we propose a decoupled and fine-grained latency regularization, that effectively overcomes our observed phenomenons that the searched networks are prone to "collapsing" to low-latency yet poor-accuracy models. Moreover, we seamlessly extend FasterSeg to a new collaborative search (co-searching) framework, simultaneously searching for a teacher and a student network in the same single run. The teacher-student distillation further boosts the student model's accuracy. Experiments on popular segmentation benchmarks demonstrate the competency of FasterSeg. For example, FasterSeg can run over 30% faster than the closest manually designed competitor on Cityscapes, while maintaining comparable accuracy.

arXiv commentICLR 2020 accepted

References in corpus (12)

• Distilling the Knowledge in a Neural Network
• Neural Architecture Search with Reinforcement Learning
• ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation
• Efficient Neural Architecture Search via Parameter Sharing
• SMASH: One-Shot Model Architecture Search through HyperNetworks
• Fast-SCNN: Fast Semantic Segmentation Network
• Deep High-Resolution Representation Learning for Visual Recognition
• Searching for Efficient Multi-Scale Architectures for Dense Image Prediction
• Slimmable Neural Networks
• Deep High-Resolution Representation Learning for Human Pose Estimation
• Joint Neural Architecture Search and Quantization
• InstaNAS: Instance-aware Neural Architecture Search

Cited by in corpus (19)

• SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers
• Deep Dual-resolution Networks for Real-time and Accurate Semantic Segmentation of Road Scenes
• GhostNets on Heterogeneous Devices via Cheap Operations
• GhostNet: More Features from Cheap Operations
• Rethinking BiSeNet For Real-time Semantic Segmentation
• Cream of the Crop: Distilling Prioritized Paths For One-Shot Neural Architecture Search
• HW-NAS-Bench:Hardware-Aware Neural Architecture Search Benchmark
• DCNAS: Densely Connected Neural Architecture Search for Semantic Image Segmentation
• Neural Feature Search for RGB-Infrared Person Re-Identification
• AutoSpeech: Neural Architecture Search for Speaker Recognition
• FBNetV5: Neural Architecture Search for Multiple Tasks in One Run
• Rethinking Dilated Convolution for Real-time Semantic Segmentation
• Understanding and Accelerating Neural Architecture Search with Training-Free and Theory-Grounded Metrics
• RARTS: An Efficient First-Order Relaxed Architecture Search Method
• Auto-Agent-Distiller: Towards Efficient Deep Reinforcement Learning Agents via Neural Architecture Search
• TriangleNet: Edge Prior Augmented Network for Semantic Segmentation through Cross-Task Consistency
• Neural Architecture Search as Sparse Supernet
• AutoBSS: An Efficient Algorithm for Block Stacking Style Search
• AASeg: Attention Aware Network for Real Time Semantic Segmentation