Keywords: Computer vision, Automl. X-ray, Security application, Neuroevolution, Neural Architecture Search, NAS
TL;DR: We use Neuroevolution with performance proxy scores to speed up Neural Architecture Search and apply it to discover state-of-the-art architecture for concealed weapon detection.
Abstract: X-ray screening is crucial for ensuring safety and security in crowded public areas. However, X-ray operators are often overwhelmed by the sheer amount of potential threats to assess; thus, current computer vision-aided systems are designed to alleviate these workloads. In this study, we focus on a key, unresolved challenge for developing such automatic X-ray screening systems: the direct application of existing avant garde computer vision approaches does not necessarily yield satisfactory results in the X-ray medium, hindering the effectiveness of current screening systems. To overcome this drawback, we propose a novel automated machine learning (AutoML) multi-objective approach for neural architecture search (NAS) for concealed weapon detection (MEOW). We benchmark MEOW with the state-of-the-art in two comprehensive scenarios in threat identification: SIXray (a popular, massive X-ray dataset) and Residuals (a proprietary, unpublished dataset provided by our industry partners). MEOW consist of the coalescence of two new components: First, we design a heuristic technique to strongly reduce the high computational cost of neuroevolutionary search while preserving a high performance such that it can be effectively used in real-time industrial settings. Second, we devise a novel ensemble approach for combining multiple discovered architectures simultaneously. Leveraging these two characteristics, MEOW outperforms the state-of-the-art while keeping the NAS overhead to a minimum. More broadly, our results suggest that AutoML has a strong potential for security applications.
Submission Checklist: Yes
Broader Impact Statement: Yes
Paper Availability And License: Yes
Code Of Conduct: Yes
Reviewers: Yes
CPU Hours: 1
GPU Hours: 1
TPU Hours: 0
Evaluation Metrics: Yes
Code And Dataset Supplement: zip
Steps For Environmental Footprint Reduction During Development: Energy-efficient hardware: We utilise energy-efficient computing hardware for model training and deployment, including energy-saving processors and GPUs designed to minimize power consumption.
Cloud-based services: We leverage cloud-based infrastructure and services for storage, computation, and deployment, which can provide optimized resource management and reduce the overall energy consumption.
Efficient algorithms: Our main focus is that MEOW is computationally efficient which makes it energy-efficient. Towards this goal, we used heuristic technique and ensemble approach in MEOW to minimize computational overhead, consequently reducing the energy consumption during model training and deployment.
Dataset optimization: Preprocess and optimize the SIXray and Residuals datasets to reduce storage requirements and computational demands during model training, allowing for more efficient use of resources.
Remote collaboration: Encourage remote collaboration among the research team and industry partners to minimize travel and the associated carbon emissions.
Estimated CO2e Footprint: 0.15
10 Replies
Loading