Angular Visual Hardness
This work addresses the calibration issue in CNNs for computer vision tasks, offering a novel metric to improve model reliability, though it is incremental as it builds on existing findings about sample hardness.
The paper tackles the problem of poor calibration in convolutional neural networks (CNNs), where models are overconfident and fail to reflect true sample hardness, by proposing angular visual hardness (AVH) as a score to measure sample hardness based on angular distance between feature embeddings and classifiers. They validate AVH by showing it correlates with model accuracy, training dynamics, human visual hardness, and benefits applications like self-training for domain adaptation and generalization.
Recent convolutional neural networks (CNNs) have led to impressive performance but often suffer from poor calibration. They tend to be overconfident, with the model confidence not always reflecting the underlying true ambiguity and hardness. In this paper, we propose angular visual hardness (AVH), a score given by the normalized angular distance between the sample feature embedding and the target classifier to measure sample hardness. We validate this score with an in-depth and extensive scientific study, and observe that CNN models with the highest accuracy also have the best AVH scores. This agrees with an earlier finding that state-of-art models improve on the classification of harder examples. We observe that the training dynamics of AVH is vastly different compared to the training loss. Specifically, AVH quickly reaches a plateau for all samples even though the training loss keeps improving. This suggests the need for designing better loss functions that can target harder examples more effectively. We also find that AVH has a statistically significant correlation with human visual hardness. Finally, we demonstrate the benefit of AVH to a variety of applications such as self-training for domain adaptation and domain generalization.