Scalable 3D Semantic Segmentation for Gun Detection in CT Scans

TL;DR

This paper presents a scalable 3D semantic segmentation method for gun detection in baggage CT scans, addressing memory and computational challenges of high-resolution 3D data with a novel moving pyramid approach.

Contribution

A new deep 3D segmentation technique that enables efficient training and inference on high-resolution volumes for gun detection in CT scans.

Findings

Enables fast training on high-resolution 3D data

Reduces memory consumption compared to existing methods

Achieves accurate gun detection in baggage CT scans

Abstract

With the increased availability of 3D data, the need for solutions processing those also increased rapidly. However, adding dimension to already reliably accurate 2D approaches leads to immense memory consumption and higher computational complexity. These issues cause current hardware to reach its limitations, with most methods forced to reduce the input resolution drastically. Our main contribution is a novel deep 3D semantic segmentation method for gun detection in baggage CT scans that enables fast training and low video memory consumption for high-resolution voxelized volumes. We introduce a moving pyramid approach that utilizes multiple forward passes at inference time for segmenting an instance.