NumbOD: A Spatial-Frequency Fusion Attack Against Object Detectors
Authors: Ziqi Zhou, Bowen Li, Yufei Song, Zhifei Yu, Shengshan Hu, Wei Wan, Leo Yu Zhang, Dezhong Yao, Hai Jin
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our extensive experiments on nine ODs and two datasets show that Numb OD achieves powerful attack performance and high stealthiness. We conduct experiments on nine object detectors and two datasets to evaluate the effectiveness of Numb OD. Both qualitative and quantitative results show that Numb OD effectively deceives ODs of various architectures, exhibiting strong attack performance and high stealthiness. |
| Researcher Affiliation | Academia | 1National Engineering Research Center for Big Data Technology and System 2Services Computing Technology and System Lab 3Cluster and Grid Computing Lab 4Hubei Engineering Research Center on Big Data Security 5Hubei Key Laboratory of Distributed System Security 6 School of Computer Science and Technology, Huazhong University of Science and Technology 7 School of Cyber Science and Engineering, Huazhong University of Science and Technology 8 School of Information and Communication Technology, Griffith University EMAIL EMAIL |
| Pseudocode | No | The paper describes the methodology in narrative text and mathematical formulas but does not include any clearly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code https://github.com/CGCL-codes/Numb OD |
| Open Datasets | Yes | For a comprehensive evaluation, we use MS-COCO (Lin et al. 2014) and PASCAL VOC (Everingham et al. 2010) datasets |
| Dataset Splits | No | The paper mentions using MS-COCO and PASCAL VOC datasets and selecting 5000 images for attack evaluation, but does not explicitly provide the training/validation/test splits used for these datasets or for the object detectors themselves. |
| Hardware Specification | No | The computation is completed in the HPC Platform of Huazhong University of Science and Technology. |
| Software Dependencies | No | The paper mentions using the MMDetection repository for fine-tuning object detectors but does not provide specific version numbers for MMDetection or any other software dependencies. |
| Experiment Setup | Yes | We set the hyperparameters λ to 100, while the training epoch is set to 50 with a batch size of 1. We utilize the Adamax optimizer and set the learning rate and weight decay to 0.03 and 0.02, respectively. |