Background adaptive faster R-CNN for semi-supervised convolutional object detection of threats in x-ray images

Abstract

Recently, progress has been made in the supervised training of Convolutional Object Detectors (e.g. Faster R-CNN) for threat recognition in carry-on luggage using X-ray images. This is part of the Transportation Security Administration's (TSA's) mission to ensure safety for air travelers in the United States. Collecting more data reliably improves performance for this class of deep algorithm, but requires time and money to produce training data with threats staged in realistic contexts. In contrast to these hand-collected data containing threats, data from the real-world, known as the Stream-of-Commerce (SOC), can be collected quickly with minimal cost; while technically unlabeled, in this work we make a practical assumption that these are without threat objects. Because of these data constraints, we will use both labeled and unlabeled sources of data for the automatic threat recognition problem. In this paper, we present a semi-supervised approach for this problem which we call Background Adaptive Faster R-CNN. This approach is a training method for two-stage object detectors which uses Domain Adaptation methods from the field of deep learning. The data sources described earlier are considered two "domains": one a hand-collected data domain of images with threats, and the other a real-world domain of images assumed without threats. Two domain discriminators, one for discriminating object proposals and one for image features, are adversarially trained to prevent encoding domain-specific information. Penalizing this encoding is important because otherwise the Convolutional Neural Network (CNN) can learn to distinguish images from the two sources based on superficial characteristics, and minimize a purely supervised loss function without improving its ability to recognize objects. For the hand-collected data, only object proposals and image features completely outside of areas corresponding to ground truth object bounding boxes (background) are used. The losses for these domain-adaptive discriminators are added to the Faster R-CNN losses of images from both domains. This technique enables threat recognition based on examples from the labeled data, and can reduce false alarm rates by matching the statistics of extracted features on the hand-collected backgrounds to that of the real world data. Performance improvements are demonstrated on two independently-collected datasets of labeled threats.

Zero shot domain adaptation for x-ray baggage scanning

“…Inspired by GAN, the concept behind UDA is to migrate the domain shift by playing an adversarial min-max zero-sum game, to force the feature extractor to learn indistinguishable features between domains. The most common adversarial training paradigm is to add a domain discriminator to the network and the min-max situation is achieved by using a gradient reversal layer (GRL) [10,16,17,18]. One limitation is that this method mainly focuses on domain level invariant features, and neglects the class level feature alignment.…”

“…Our training structure follows the end-to-end UDA approach proposed by Ganin, et. al [10] with our own modifications. As shown in Figure 2, the model contains three components.…”

“…For instance, images obtained from different X-ray scanners with different hardware intrinsics are often not represented in synthetic data [9]. To address these domain gap problems, an adversarial domain adaptation method has been introduced recently to jointly train an ATR model between domains [10]. This method follows the general unsupervised domain adaptation (UDA) approach first proposed by Ganin.…”

Methods and Technologies for Detecting Potentially Dangerous Objects in a Video Stream From Introscope

“…В статье [3] используется модифицированная версия архитектуры Faster R-CNN с целью улучшения детекции объектов на изображениях путем учета и адаптации к сложному фону. Эта модификация призвана бороться с проблемой, когда объекты могут затеряться на сложном фоне, что может привести к неверным детекциям или потере объектов.…”

Multi-Object Detection in Security Screening Scene Based on Convolutional Neural Network

“…Literature [ 27 ] suggests a lateral inhibition module (LIM) that maximally inhibits the flow of noisy information via a bidirectional propagation (BP) module and activates the most attractive borders from four directions via a boundary activation (BA) module. Literature [ 28 ] employs adversarial domain adaption approaches to match the background distribution of a large number of unlabeled SoC samples. This approach helps to train the network to detect items in the SoC dataset using a small labeled dataset.…”