Token Turing Machines are Efficient Vision Models

TL;DR

The paper introduces Vision Token Turing Machines (ViTTM), a memory-augmented transformer that significantly reduces inference time and computational cost for vision tasks while maintaining or improving accuracy.

Contribution

It presents a novel architecture combining process and memory tokens for efficient, low-latency vision modeling, extending Turing machine concepts to vision transformers.

Findings

ViTTM-B is 56% faster than ViT-B on ImageNet-1K.

ViTTM-B achieves higher accuracy (82.9%) with fewer FLOPs.

On ADE20K, ViTTM-B doubles the FPS compared to ViT-B while maintaining similar mIoU.

Abstract

We propose Vision Token Turing Machines (ViTTM), an efficient, low-latency, memory-augmented Vision Transformer (ViT). Our approach builds on Neural Turing Machines and Token Turing Machines, which were applied to NLP and sequential visual understanding tasks. ViTTMs are designed for non-sequential computer vision tasks such as image classification and segmentation. Our model creates two sets of tokens: process tokens and memory tokens; process tokens pass through encoder blocks and read-write from memory tokens at each encoder block in the network, allowing them to store and retrieve information from memory. By ensuring that there are fewer process tokens than memory tokens, we are able to reduce the inference time of the network while maintaining its accuracy. On ImageNet-1K, the state-of-the-art ViT-B has median latency of 529.5ms and 81.0% accuracy, while our ViTTM-B is 56% faster (234.1ms), with 2.4 times fewer FLOPs, with an accuracy of 82.9%. On ADE20K semantic segmentation, ViT-B achieves 45.65mIoU at 13.8 frame-per-second (FPS) whereas our ViTTM-B model acheives a 45.17 mIoU with 26.8 FPS (+94%).