Papers › Revisiting ResNets: Improved Training and Scaling Strategies

Revisiting ResNets: Improved Training and Scaling Strategies

13 Mar 2021NeurIPS 2021 12arXiv:2103.07579archive 2025-07-28

Irwan Bello, William Fedus, Xianzhi Du, Ekin D. Cubuk, Aravind Srinivas, Tsung-Yi Lin, Jonathon Shlens, Barret Zoph

Novel computer vision architectures monopolize the spotlight, but the impact of the model architecture is often conflated with simultaneous changes to training methodology and scaling strategies. Our work revisits the canonical ResNet (He et al., 2015) and studies these three aspects in an effort to disentangle them. Perhaps surprisingly, we find that training and scaling strategies may matter more than architectural changes, and further, that the resulting ResNets match recent state-of-the-art models. We show that the best performing scaling strategy depends on the training regime and offer two new scaling strategies: (1) scale model depth in regimes where overfitting can occur (width scaling is preferable otherwise); (2) increase image resolution more slowly than previously recommended (Tan & Le, 2019). Using improved training and scaling strategies, we design a family of ResNet architectures, ResNet-RS, which are 1.7x - 2.7x faster than EfficientNets on TPUs, while achieving similar accuracies on ImageNet. In a large-scale semi-supervised learning setup, ResNet-RS achieves 86.2% top-1 ImageNet accuracy, while being 4.7x faster than EfficientNet NoisyStudent. The training techniques improve transfer performance on a suite of downstream tasks (rivaling state-of-the-art self-supervised algorithms) and extend to video classification on Kinetics-400. We recommend practitioners use these simple revised ResNets as baselines for future research.

PaperPDFConference PDFCode

In Syntology View this paper on Syntology, its page in Syntology's graph. That page lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

tensorflow/tpu officialtfSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Towards Understanding Why Label Smoothing Degrades Selective Classification and How to Fix It (arXiv:2403.14715; that paper's own run record: 9 ran (of which 2 constructed an object rather than computing a result; 6 with no instrument failure: 1 honoured, 0 violated, 5 with no contract checked; 3 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method. report
nachiket273/pytorch_resnet_rs mentioned on GitHubpytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Deep Networks with Stochastic Depth (arXiv:1603.09382; that paper's own run record: 8 ran (of which 5 constructed an object rather than computing a result; 7 with no instrument failure: 0 honoured, 1 violated, 6 with no contract checked; 1 where Syntology's instrument failed) · 1 unverified) and RandAugment: Practical automated data augmentation with a reduced search space (arXiv:1909.13719; that paper's own run record: 58 ran (of which 1 constructed an object rather than computing a result; 7 with no instrument failure: 0 honoured, 1 violated, 6 with no contract checked; 51 where Syntology's instrument failed) · 7 unverified); nothing checks that they implement this paper's method. report
rwightman/pytorch-image-models mentioned on GitHubpytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to AdamP: Slowing Down the Slowdown for Momentum Optimizers on Scale-invariant Weights (arXiv:2006.08217; that paper's own run record: 4 ran (of which 1 constructed an object rather than computing a result; 2 with no instrument failure: 1 honoured, 0 violated, 1 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method. report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Action ClassificationDocument Image ClassificationImage ClassificationSemantic Object Interaction ClassificationVideo Classification

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Document Image Classification AIP ResNet-RS (ResNet-200 + RS training tricks) Top 1 Accuracy - Verb 83.4 #1 of 1 Archive leaderboard report
Image Classification ImageNet ResNet-RS-50 (160 image res) GFLOPs 4.6 #322 of 1060 Archive leaderboard report
Image Classification ImageNet ResNet-RS-50 (160 image res) Number of params 192M #322 of 1060 Archive leaderboard report
Image Classification ImageNet ResNet-RS-50 (160 image res) Top 1 Accuracy 84.4% #322 of 1060 Archive leaderboard report
Image Classification ImageNet ResNet-RS-270 (256 image res) GFLOPs 54 #384 of 1060 Archive leaderboard report
Image Classification ImageNet ResNet-RS-270 (256 image res) Top 1 Accuracy 83.8% #384 of 1060 Archive leaderboard report
Image Classification PRImA ResNet-152 2x (RS training) Percentage correct 89.3 #1 of 1 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Introduced by this paper: ResNet-RS

1x1 ConvolutionAverage PoolingBatch NormalizationBottleneck Residual BlockConvolutionCosine AnnealingDense ConnectionsDepthwise ConvolutionDepthwise Separable ConvolutionDropoutEfficientNetGlobal Average PoolingInverted Residual BlockKaiming InitializationLabel SmoothingMax PoolingPointwise ConvolutionRMSPropReLUResNet-DResNet-RSResidual BlockResidual ConnectionSigmoid ActivationSqueeze-and-Excitation BlockStochastic DepthWeight DecayXavier Initialization

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections