Testing Deep Learning Models for Image Analysis Using Object-Relevant Metamorphic Relations
This addresses concerns about model reliability in image analysis for users in fields like computer vision and AI safety, though it is incremental as it builds on existing metamorphic testing methods.
The paper tackled the problem of deep learning models making inferences based on irrelevant features in image analysis, proposing a metamorphic testing approach that revealed over 5.3% of top-5 correct predictions in image classification models and over 8.5% in object detection models are subject to such inappropriate inferences.
Deep learning models are widely used for image analysis. While they offer high performance in terms of accuracy, people are concerned about if these models inappropriately make inferences using irrelevant features that are not encoded from the target object in a given image. To address the concern, we propose a metamorphic testing approach that assesses if a given inference is made based on irrelevant features. Specifically, we propose two novel metamorphic relations to detect such inappropriate inferences. We applied our approach to 10 image classification models and 10 object detection models, with three large datasets, i.e., ImageNet, COCO, and Pascal VOC. Over 5.3% of the top-5 correct predictions made by the image classification models are subject to inappropriate inferences using irrelevant features. The corresponding rate for the object detection models is over 8.5%. Based on the findings, we further designed a new image generation strategy that can effectively attack existing models. Comparing with a baseline approach, our strategy can double the success rate of attacks.