Revisiting Multimodal Fusion for 3D Anomaly Detection from an Architectural Perspective
Authors: Kaifang Long, Guoyang Xie, Lianbo Ma, Jiaqi Liu, Zhichao Lu
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments show that 3D-ADNAS obtains consistent improvements in 3D-AD across various model capacities in terms of accuracy, frame rate, and memory usage, and it exhibits great potential in dealing with few-shot 3D-AD tasks. ... We adopt empirical and theoretical approaches and conduct extensive experiments to realize this goal. ... Extensive experiments validate the effectiveness of our design in improving anomaly detection of multimodal fusion network in terms of accuracy, speed and efficiency. |
| Researcher Affiliation | Collaboration | Kaifang Long1*, Guoyang Xie2*, Lianbo Ma1 , Jiaqi Liu3, Zhichao Lu3 1Software College, Northeastern University, Shenyang, China 2 The Department of Intelligent Manufacturing, CATL, Ningde, China 3The Department of Computer Science, City University of Hong Kong, Hong Kong, China EMAIL, EMAIL, EMAIL, liu jiaqi @outlook.com, EMAIL |
| Pseudocode | No | The paper describes algorithms and methods but does not present any structured pseudocode or algorithm blocks with explicit labels like 'Algorithm' or 'Pseudocode'. |
| Open Source Code | Yes | More detailed proofs are provided in the Supplementary Material1. 1Please refer to the following for additional material: https://github.com/longkaifang/3D-ADNAS. |
| Open Datasets | Yes | We evaluate 3D-ADNAS on Eyecandies (Bonfiglioli et al. 2022) and MVTec 3D-AD (Bergmann et al. 2022) datasets. |
| Dataset Splits | Yes | To evaluate the effectiveness of 3D-ADNAS in few-shot scenarios, we randomly select 5, 10, and 50 images on Eyecandies and MVTec 3D-AD datasets as the training set and perform inference on the full test set. |
| Hardware Specification | Yes | As shown in Table 5, we can see that 3D-ADNAS gets fastest frame rate, highest I-AUROC scores, and lowest memory usage when tested on single NVIDIA RTX 4090. |
| Software Dependencies | No | The paper mentions using the 'Adam optimizer' and following 'the setting of DARTS' but does not specify any software versions for libraries like PyTorch, TensorFlow, or Python. |
| Experiment Setup | Yes | For the tests about impact of intra/inter-module fusion architectures, we train the overall model of 3D-ADNAS by 600 epochs. For the tests about evaluation of searched architectures, we train MFN model by 80 epochs and follow the setting of DARTS to obtain best multimodal fusion model, and then train the overall model of 3D-ADNAS with obtained MFN by 600 epochs. ... We train the 3D-ADNAS with the Adam optimizer. |