Heterogeneous Generative Knowledge Distillation with Masked Image Modeling

TL;DR

This paper introduces Heterogeneous Generative Knowledge Distillation (H-GKD), a novel method that transfers knowledge from large Transformer models to small CNNs using masked image modeling, improving performance across multiple visual tasks.

Contribution

It is the first to apply masked image modeling for knowledge distillation between heterogeneous models, bridging Transformer and CNN architectures effectively.

Findings

Achieves state-of-the-art results in image classification, object detection, and segmentation.

Improves Resnet50 accuracy from 76.98% to 80.01% on ImageNet 1K.

Demonstrates robustness across various model sizes and tasks.

Abstract

Small CNN-based models usually require transferring knowledge from a large model before they are deployed in computationally resource-limited edge devices. Masked image modeling (MIM) methods achieve great success in various visual tasks but remain largely unexplored in knowledge distillation for heterogeneous deep models. The reason is mainly due to the significant discrepancy between the Transformer-based large model and the CNN-based small network. In this paper, we develop the first Heterogeneous Generative Knowledge Distillation (H-GKD) based on MIM, which can efficiently transfer knowledge from large Transformer models to small CNN-based models in a generative self-supervised fashion. Our method builds a bridge between Transformer-based models and CNNs by training a UNet-style student with sparse convolution, which can effectively mimic the visual representation inferred by a teacher over masked modeling. Our method is a simple yet effective learning paradigm to learn the visual representation and distribution of data from heterogeneous teacher models, which can be pre-trained using advanced generative methods. Extensive experiments show that it adapts well to various models and sizes, consistently achieving state-of-the-art performance in image classification, object detection, and semantic segmentation tasks. For example, in the Imagenet 1K dataset, H-GKD improves the accuracy of Resnet50 (sparse) from 76.98% to 80.01%.