Fast-SCNN: Fast Semantic Segmentation Network
This addresses the need for efficient, real-time segmentation on embedded devices in autonomous systems, representing an incremental improvement over existing two-branch methods.
The paper tackles real-time semantic segmentation for autonomous systems by introducing Fast-SCNN, a model that achieves 68.0% mean intersection over union at 123.5 frames per second on high-resolution Cityscapes images, without requiring large-scale pre-training.
The encoder-decoder framework is state-of-the-art for offline semantic image segmentation. Since the rise in autonomous systems, real-time computation is increasingly desirable. In this paper, we introduce fast segmentation convolutional neural network (Fast-SCNN), an above real-time semantic segmentation model on high resolution image data (1024x2048px) suited to efficient computation on embedded devices with low memory. Building on existing two-branch methods for fast segmentation, we introduce our `learning to downsample' module which computes low-level features for multiple resolution branches simultaneously. Our network combines spatial detail at high resolution with deep features extracted at lower resolution, yielding an accuracy of 68.0% mean intersection over union at 123.5 frames per second on Cityscapes. We also show that large scale pre-training is unnecessary. We thoroughly validate our metric in experiments with ImageNet pre-training and the coarse labeled data of Cityscapes. Finally, we show even faster computation with competitive results on subsampled inputs, without any network modifications.