# Masked Image Residual Learning for Scaling Deeper Vision Transformers

Source: [Masked Image Residual Learning for Scaling Deeper Vision Transformers](<https://www.sota2.com/research/paper/68fa71291e209131cea4f93d>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-04T03:53:00.782Z

Authors: Guoxi Huang, Hongtao Fu, Adrian G. Bors

Published: 2023-09-25T13:45:28.000Z

## Abstract

Deeper Vision Transformers (ViTs) are more challenging to train. We expose a degradation problem in deeper layers of ViT when using masked image modeling (MIM) for pre-training. To ease the training of deeper ViTs, we introduce a self-supervised learning framework called Masked Image Residual Learning (MIRL), which significantly alleviates the degradation problem, making scaling ViT along depth a promising direction for performance upgrade. We reformulate the pre-training objective for deeper layers of ViT as learning to recover the residual of the masked image. We provide extensive empirical evidence showing that deeper ViTs can be effectively optimized using MIRL and easily gain accuracy from increased depth. With the same level of computational complexity as ViT-Base and ViT-Large, we instantiate 4.5$\\times$ and 2$\\times$ deeper ViTs, dubbed ViT-S-54 and ViT-B-48. The deeper ViT-S-54, costing 3$\\times$ less than ViT-Large, achieves performance on par with ViT-Large. ViT-B-48 achieves 86.2% top-1 accuracy on ImageNet. On one hand, deeper ViTs pre-trained with MIRL exhibit excellent generalization capabilities on downstream tasks, such as object detection and semantic segmentation. On the other hand, MIRL demonstrates high pre-training efficiency. With less pre-training time, MIRL yields competitive performance compared to other approaches.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Image Classification on ImageNet 1K (val) (FT Accuracy)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-1k-val-ft-accuracy.md>): Image Classification; ImageNet-1k (val); FT Accuracy: 86.2.

- [Image Classification on ImageNet-1K (Fine-tuning)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-1k-fine-tuning.md>): Image Classification; ImageNet-1K (fine-tuning); Accuracy (FT): 84.1.

- [Semantic Segmentation on ADE20K](<https://www.sota2.com/research/sota/semantic-segmentation-on-ade20k.md>): Semantic segmentation; ADE20K; mIoU: 53.2.

- [Instance Segmentation on COCO 2017 (val)](<https://www.sota2.com/research/sota/instance-segmentation-on-coco-2017-val.md>): Instance Segmentation; COCO 2017 (val); APm: 0.437.

- [Object Detection on COCO 2017 (val)](<https://www.sota2.com/research/sota/object-detection-on-coco-2017-val.md>): Object Detection; COCO 2017 (val); AP: not recorded.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2309.14136>)

[PDF](<https://arxiv.org/pdf/2309.14136.pdf>)

[Implementation](<https://github.com/russellllaputa/MIRL>)
