Atrous Space Bender U-Net (ASBU-Net/LogiNet)

TL;DR

ASBU-Net is a fast, efficient CNN for high-resolution image segmentation that uses a novel atrous space bender layer, enabling easy deployment on hardware with limited memory while maintaining high accuracy.

Contribution

The paper introduces ASBU-Net with a new atrous space bender layer, improving efficiency and hardware compatibility for semantic segmentation tasks.

Findings

Strong performance on resource and accuracy trade-offs

Compatible with FPGA and embedded hardware

No special layers needed for quantization

Abstract

$ $With recent advances in CNNs, exceptional improvements have been made in semantic segmentation of high resolution images in terms of accuracy and latency. However, challenges still remain in detecting objects in crowded scenes, large scale variations, partial occlusion, and distortions, while still maintaining mobility and latency. We introduce a fast and efficient convolutional neural network, ASBU-Net, for semantic segmentation of high resolution images that addresses these problems and uses no novelty layers for ease of quantization and embedded hardware support. ASBU-Net is based on a new feature extraction module, atrous space bender layer (ASBL), which is efficient in terms of computation and memory. The ASB layers form a building block that is used to make ASBNet. Since this network does not use any special layers it can be easily implemented, quantized and deployed on FPGAs and other hardware with limited memory. We present experiments on resource and accuracy trade-offs and show strong performance compared to other popular models.