Salience-Based Adaptive Masking: Revisiting Token Dynamics for Enhanced Pre-training
This work addresses a key bottleneck in MIM-based pre-training for computer vision, offering a cost-effective solution to improve model performance and stability.
The paper tackles the problem of performance instability in Masked Image Modeling (MIM) pre-training by introducing Saliency-Based Adaptive Masking (SBAM), which prioritizes token salience to enhance robustness against masking ratio variations, resulting in significant improvements over state-of-the-art methods on the ImageNet-1K dataset.
In this paper, we introduce Saliency-Based Adaptive Masking (SBAM), a novel and cost-effective approach that significantly enhances the pre-training performance of Masked Image Modeling (MIM) approaches by prioritizing token salience. Our method provides robustness against variations in masking ratios, effectively mitigating the performance instability issues common in existing methods. This relaxes the sensitivity of MIM-based pre-training to masking ratios, which in turn allows us to propose an adaptive strategy for `tailored' masking ratios for each data sample, which no existing method can provide. Toward this goal, we propose an Adaptive Masking Ratio (AMR) strategy that dynamically adjusts the proportion of masking for the unique content of each image based on token salience. We show that our method significantly improves over the state-of-the-art in mask-based pre-training on the ImageNet-1K dataset.