Papers › SpineNet: Learning Scale-Permuted Backbone for Recognition and Localization

SpineNet: Learning Scale-Permuted Backbone for Recognition and Localization

10 Dec 2019CVPR 2020 6arXiv:1912.05027archive 2025-07-28

Xianzhi Du, Tsung-Yi Lin, Pengchong Jin, Golnaz Ghiasi, Mingxing Tan, Yin Cui, Quoc V. Le, Xiaodan Song

Convolutional neural networks typically encode an input image into a series of intermediate features with decreasing resolutions. While this structure is suited to classification tasks, it does not perform well for tasks requiring simultaneous recognition and localization (e.g., object detection). The encoder-decoder architectures are proposed to resolve this by applying a decoder network onto a backbone model designed for classification tasks. In this paper, we argue encoder-decoder architecture is ineffective in generating strong multi-scale features because of the scale-decreased backbone. We propose SpineNet, a backbone with scale-permuted intermediate features and cross-scale connections that is learned on an object detection task by Neural Architecture Search. Using similar building blocks, SpineNet models outperform ResNet-FPN models by ~3% AP at various scales while using 10-20% fewer FLOPs. In particular, SpineNet-190 achieves 52.5% AP with a MaskR-CNN detector and achieves 52.1% AP with a RetinaNet detector on COCO for a single model without test-time augmentation, significantly outperforms prior art of detectors. SpineNet can transfer to classification tasks, achieving 5% top-1 accuracy improvement on a challenging iNaturalist fine-grained dataset. Code is at: https://github.com/tensorflow/tpu/tree/master/models/official/detection.

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

13 repositories listed; official and paper-mentioned ones first.

tensorflow/tpu officialmentioned in papertfApache-2.0Syntology: 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
tensorflow/models officialtfSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift (arXiv:1502.03167; that paper's own run record: 19 ran (of which 5 constructed an object rather than computing a result; 14 with no instrument failure: 0 honoured, 0 violated, 14 with no contract checked; 5 where Syntology's instrument failed) · 2 unverified), Deep Residual Learning for Image Recognition (arXiv:1512.03385; that paper's own run record: 254 ran (of which 108 constructed an object rather than computing a result; 163 with no instrument failure: 3 honoured, 1 violated, 159 with no contract checked; 91 where Syntology's instrument failed) · 123 unverified), Wide Residual Networks (arXiv:1605.07146; that paper's own run record: 66 ran (of which 31 constructed an object rather than computing a result; 47 with no instrument failure: 0 honoured, 0 violated, 47 with no contract checked; 19 where Syntology's instrument failed) · 30 unverified) and 1 more (every one in this page's JSON twin); nothing checks that they implement this paper's method. report
lucifer443/SpineNet-Pytorch mentioned on GitHubpytorchApache-2.0Syntology: not harvested report
tensorflow/tpu mentioned on GitHubtfSyntology: 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
yan-roo/SpineNet-Pytorch mentioned on GitHubpytorchApache-2.0Syntology: not harvested report
2023-MindSpore-4/Code11 mindsporeSyntology: not harvested report
2023-MindSpore-4/Code14 mindsporeSyntology: not harvested report
2023-MindSpore-4/Code14 mindsporeSyntology: not harvested report
2023-MindSpore-4/Code14 mindsporeSyntology: not harvested report
2023-MindSpore-4/Code9 mindsporeSyntology: not harvested report
MindSpore-paper-code-2/code3 mindsporeSyntology: not harvested report
code-implementation1/Code7 mindsporeSyntology: not harvested report
code-implementation1/Code8 mindsporeSyntology: not harvested 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

Image ClassificationInstance SegmentationNeural Architecture SearchObject DetectionPoint Cloud RegistrationReal-Time Object Detectionobject-detection

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Classification ImageNet SpineNet-143 GFLOPs 9.1 #794 of 1060 Archive leaderboard report
Image Classification ImageNet SpineNet-143 Number of params 60.5M #794 of 1060 Archive leaderboard report
Image Classification ImageNet SpineNet-143 Top 1 Accuracy 79% #794 of 1060 Archive leaderboard report
Image Classification iNaturalist SpineNet-143 Top 1 Accuracy 63.6% #15 of 19 Archive leaderboard report
Image Classification iNaturalist SpineNet-143 Top 5 Accuracy 84.8% #15 of 19 Archive leaderboard report
Instance Segmentation COCO minival RetinaNet (SpineNet-190, 1536x1536) mask AP 46.1 #45 of 93 Archive leaderboard report
Instance Segmentation COCO test-dev Mask R-CNN (SpineNet-190, 1536x1536) mask AP 46.1 #38 of 112 Archive leaderboard report
Object Detection COCO minival RetinaNet (SpineNet-190, 1536x1536) box AP 52.2 #67 of 220 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-190, 1280x1280) AP50 71.8 #74 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-190, 1280x1280) AP75 56.5 #74 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-190, 1280x1280) APL 63.6 #74 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-190, 1280x1280) APM 55 #74 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-190, 1280x1280) APS 35.4 #74 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-190, 1280x1280) box mAP 52.1 #74 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-143, 1280x1280) AP50 70.4 #85 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-143, 1280x1280) AP75 54.9 #85 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-143, 1280x1280) APL 62.1 #85 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-143, 1280x1280) APM 53.9 #85 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-143, 1280x1280) APS 33.6 #85 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-143, 1280x1280) box mAP 50.7 #85 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-96, 1024x1024) AP50 68.4 #104 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-96, 1024x1024) AP75 52.5 #104 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-96, 1024x1024) APL 62 #104 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-96, 1024x1024) APM 52.3 #104 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-96, 1024x1024) APS 32 #104 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-96, 1024x1024) box mAP 48.6 #104 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49, 896x896) AP50 66.3 #124 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49, 896x896) AP75 50.6 #124 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49, 896x896) APL 61.7 #124 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49, 896x896) APM 50.1 #124 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49, 896x896) APS 29.1 #124 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49, 896x896) box mAP 46.7 #124 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49, 640x640) AP50 63.8 #146 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49, 640x640) AP75 47.6 #146 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49, 640x640) APL 61.1 #146 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49, 640x640) APM 47.7 #146 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49, 640x640) APS 25.9 #146 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49, 640x640) box mAP 44.3 #146 of 225 Archive leaderboard report
Object Detection COCO test-dev SpineNet-49 (640, RetinaNet, single-scale) AP50 62.3 #167 of 225 Archive leaderboard report
Object Detection COCO test-dev SpineNet-49 (640, RetinaNet, single-scale) AP75 46.1 #167 of 225 Archive leaderboard report
Object Detection COCO test-dev SpineNet-49 (640, RetinaNet, single-scale) APL 57.3 #167 of 225 Archive leaderboard report
Object Detection COCO test-dev SpineNet-49 (640, RetinaNet, single-scale) APM 45.2 #167 of 225 Archive leaderboard report
Object Detection COCO test-dev SpineNet-49 (640, RetinaNet, single-scale) APS 23.7 #167 of 225 Archive leaderboard report
Object Detection COCO test-dev SpineNet-49 (640, RetinaNet, single-scale) box mAP 42.8 #167 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49S, 640x640) AP50 60.5 #183 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49S, 640x640) AP75 44.6 #183 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49S, 640x640) APL 58 #183 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49S, 640x640) APM 45 #183 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49S, 640x640) APS 23.3 #183 of 225 Archive leaderboard report
Object Detection COCO test-dev RetinaNet (SpineNet-49S, 640x640) box mAP 41.5 #183 of 225 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: SpineNet

1x1 ConvolutionAverage PoolingBatch NormalizationBottleneck Residual BlockConvolutionCosine AnnealingDense ConnectionsEntropy RegularizationFPNFocal LossGlobal Average PoolingKaiming InitializationLSTMLinear Warmup With Cosine AnnealingMask R-CNNMax PoolingNAS-FPNNeural Architecture SearchPPORPNRandom Horizontal FlipRandom Resized CropReLUResidual BlockResidual ConnectionRetinaNetRoIAlignSGD with MomentumSigmoid ActivationSoftmaxSpineNetStochastic DepthTanh ActivationWeight Decay

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