AutoFormer: Searching Transformers for Visual Recognition

Recently, pure transformer-based models have shown great potentials for vision tasks such as image classification and detection. However, the design of transformer networks is challenging... It has been observed that the depth, embedding dimension, and number of heads can largely affect the performance of vision transformers. Previous models configure these dimensions based upon manual crafting. In this work, we propose a new one-shot architecture search framework, namely AutoFormer, dedicated to vision transformer search. AutoFormer entangles the weights of different blocks in the same layers during supernet training. Benefiting from the strategy, the trained supernet allows thousands of subnets to be very well-trained. Specifically, the performance of these subnets with weights inherited from the supernet is comparable to those retrained from scratch. Besides, the searched models, which we refer to AutoFormers, surpass the recent state-of-the-arts such as ViT and DeiT. In particular, AutoFormer-tiny/small/base achieve 74.7%/81.7%/82.4% top-1 accuracy on ImageNet with 5.7M/22.9M/53.7M parameters, respectively. Lastly, we verify the transferability of AutoFormer by providing the performance on downstream benchmarks and distillation experiments. Code and models are available at https://github.com/microsoft/AutoML. read more

PDF Abstract ICCV 2021 PDF ICCV 2021 Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Result Benchmark
Image Classification CIFAR-10 AutoFormer-S | 384 Percentage correct 99.1 # 6
PARAMS 23M # 157
Image Classification CIFAR-100 AutoFormer-S | 384 Percentage correct 91.1 # 17
PARAMS 23M # 149
Image Classification ImageNet AutoFormer-base Top 1 Accuracy 82.4% # 177
Number of params 54M # 123
Image Classification ImageNet AutoFormer-tiny Top 1 Accuracy 74.7% # 423
Top 5 Accuracy 92.6 # 180
Number of params 5.7M # 267
Image Classification ImageNet AutoFormer-small Top 1 Accuracy 81.7% # 211
Top 5 Accuracy 95.7 # 79
Number of params 22.9M # 199
Fine-Grained Image Classification Oxford 102 Flowers AutoFormer-S | 384 Top 1 Accuracy 98.8 # 3
Fine-Grained Image Classification Oxford-IIIT Pets AutoFormer-S | 384 Accuracy 94.9 # 6
Fine-Grained Image Classification Stanford Cars AutoFormer-S | 384 Accuracy 93.4% # 37

Methods