Cross-Architecture Knowledge Distillation

Yufan Liu, Jiajiong Cao, Bing Li, Weiming Hu, Jingting Ding, Liang Li; Proceedings of the Asian Conference on Computer Vision (ACCV), 2022, pp. 3396-3411


Transformer attracts much attention because of its ability to learn global relations and superior performance. In order to achieve higher performance, it is natural to distill complementary knowledge from Transformer to convolutional neural network (CNN). However, most existing knowledge distillation methods only consider homologous-architecture distillation, such as distilling knowledge from CNN to CNN. They may not be suitable when applying to cross-architecture scenarios, such as from Transformer to CNN. To deal with this problem, a novel cross-architecture knowledge distillation method is proposed. Specifically, instead of directly mimicking output/intermediate features of the teacher, a partially cross attention projector and a group-wise linear projector are introduced to align the student features with the teacher's in two projected feature spaces. And a multi-view robust training scheme is further presented to improve the robustness and stability of the framework. Extensive experiments show that the proposed method outperforms 14 state-of-the-arts on both small-scale and large-scale datasets.

Related Material

[pdf] [arXiv]
@InProceedings{Liu_2022_ACCV, author = {Liu, Yufan and Cao, Jiajiong and Li, Bing and Hu, Weiming and Ding, Jingting and Li, Liang}, title = {Cross-Architecture Knowledge Distillation}, booktitle = {Proceedings of the Asian Conference on Computer Vision (ACCV)}, month = {December}, year = {2022}, pages = {3396-3411} }