Zeros can be Informative: Masked Binary U-Net for Image Segmentation on Tensor Cores

TL;DR

This paper introduces Masked Binary U-Net, a hardware-efficient binary neural network for real-time high-resolution image segmentation that maintains near full-precision accuracy while significantly improving speed and energy efficiency on GPUs.

Contribution

The paper proposes a novel masked binary U-Net architecture and a GPU implementation that together achieve high accuracy and efficiency for real-time image segmentation on resource-constrained devices.

Findings

Near full-precision accuracy with only 3% average accuracy drop.

Over 2x speedup compared to 16-bit floating point U-Net.

More than 3x energy reduction on GPUs.

Abstract

Real-time image segmentation is a key enabler for AR/VR, robotics, drones, and autonomous systems, where tight accuracy, latency, and energy budgets must be met on resource-constrained edge devices. While U-Net offers a favorable balance of accuracy and efficiency compared to large transformer-based models, achieving real-time performance on high-resolution input remains challenging due to compute, memory, and power limits. Extreme quantization, particularly binary networks, is appealing for its hardware-friendly operations. However, two obstacles limit practicality: (1) severe accuracy degradation, and (2) a lack of end-to-end implementations that deliver efficiency on general-purpose GPUs. We make two empirical observations that guide our design. (1) An explicit zero state is essential: training with zero masking to binary U-Net weights yields noticeable sparsity. (2) Quantization sensitivity is uniform across layers. Motivated by these findings, we introduce Masked Binary U-Net (MBU-Net), obtained through a cost-aware masking strategy that prioritizes masking where it yields the highest accuracy-per-cost, reconciling accuracy with near-binary efficiency. To realize these gains in practice, we develop a GPU execution framework that maps MBU-Net to Tensor Cores via a subtractive bit-encoding scheme, efficiently implementing masked binary weights with binary activations. This design leverages native binary Tensor Core BMMA instructions, enabling high throughput and energy savings on widely available GPUs. Across 3 segmentation benchmarks, MBU-Net attains near full-precision accuracy (3% average drop) while delivering 2.04x speedup and 3.54x energy reductions over a 16-bit floating point U-Net.