Object Separation for X-ray Security Images

Pipeline Framework

This repository contains the evaluation pipeline code for our paper entitled "Pixel-level Analysis for Improving Threat Detection in X-ray Security Images". We also provide the pixel-level annotations on a randomly sampled subset from the SIXray dataset as well as the list of mislabeled negative samples in the data folder.

This code was tested on an Ubuntu 18.04.5 machine with Python 3.7.11.
Please install detectron2, smp, and albumentations.

# install detectron2
git clone https://github.com/facebookresearch/detectron2.git detectron2_repo
pip install -e detectron2_repo
# install smp and albumentations
pip install -U segmentation-models-pytorch albumentations --user

To run the code:

python test_pipeline.py --img_path /path/to/image.jpg

Overlayed output mask

Classification mAP on SIXray100 subset:

Method mean Gun Knife Wrench Pliers Scissors
CHR 64.54 98.60 81.09 46.69 62.72 33.61
GBAD 66.14 95.74 85.25 54.87 53.10 41.74
Mask R-CNN 86.55 98.74 77.72 78.35 89.78 77.11
Proposed 91.40 99.46 90.24 86.88 94.37 86.06

References