Camera Bias in a Fine Grained Classification Task
This reveals a critical bias issue in image classification models that can undermine real-world applicability, particularly for tasks like fine-grained recognition.
The paper demonstrates that convolutional neural networks exploit correlations between camera models and class labels in fine-grained image classification, leading to poor generalization when those correlations are absent or with unseen cameras, and identifies high-frequency features as key for camera recognition.
We show that correlations between the camera used to acquire an image and the class label of that image can be exploited by convolutional neural networks (CNN), resulting in a model that "cheats" at an image classification task by recognizing which camera took the image and inferring the class label from the camera. We show that models trained on a dataset with camera / label correlations do not generalize well to images in which those correlations are absent, nor to images from unencountered cameras. Furthermore, we investigate which visual features they are exploiting for camera recognition. Our experiments present evidence against the importance of global color statistics, lens deformation and chromatic aberration, and in favor of high frequency features, which may be introduced by image processing algorithms built into the cameras.