{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/filter-response-normalization-layer","title":"Filter Response Normalization Layer: Eliminating Batch Dependence in the Training of Deep Neural Networks","arxiv_id":"1911.09737","date":"2019-11-21","proceeding":"CVPR 2020 6","authors":["Saurabh Singh","Shankar Krishnan"],"abstract":"Batch Normalization (BN) uses mini-batch statistics to normalize the activations during training, introducing dependence between mini-batch elements. This dependency can hurt the performance if the mini-batch size is too small, or if the elements are correlated. Several alternatives, such as Batch Renormalization and Group Normalization (GN), have been proposed to address this issue. However, they either do not match the performance of BN for large batches, or still exhibit degradation in performance for smaller batches, or introduce artificial constraints on the model architecture. In this paper we propose the Filter Response Normalization (FRN) layer, a novel combination of a normalization and an activation function, that can be used as a replacement for other normalizations and activations. Our method operates on each activation channel of each batch element independently, eliminating the dependency on other batch elements. Our method outperforms BN and other alternatives in a variety of settings for all batch sizes. FRN layer performs $\\approx 0.7-1.0\\%$ better than BN on top-1 validation accuracy with large mini-batch sizes for Imagenet classification using InceptionV3 and ResnetV2-50 architectures. Further, it performs $>1\\%$ better than GN on the same problem in the small mini-batch size regime. For object detection problem on COCO dataset, FRN layer outperforms all other methods by at least $0.3-0.5\\%$ in all batch size regimes.","url_abs":"https://arxiv.org/abs/1911.09737v2","url_pdf":"https://arxiv.org/pdf/1911.09737v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/CarloLepelaars/filter_response_normalization_keras","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/EmptySamurai/pytorch-reconet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/SdahlSean/FilterResponseNormalizationTensorflow","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/amirbar/FilterResponseNormalization","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/gakkiri/Filter-Response-Normalization","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/gupta-abhay/pytorch-frn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/kobiso/FilterResponseNorm-MXNet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"mxnet","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/philipperemy/keras-frn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/ppwwyyxx/FRN-on-common-ImageNet-baseline","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/tattaka/Filter-Response-Normalization-PyTorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/2023-MindSpore-1/ms-code-216/tree/main/resnetv2_50_frn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/2023-MindSpore-4/Code11/tree/main/resnetv2_50_frn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/2023-MindSpore-4/Code14/tree/main/resnetv2_50_frn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/MindSpore-paper-code-3/code5/tree/main/resnetv2_50_frn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/ensta-u2is/torch-uncertainty","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"filter-response-normalization-layer","repo_url":"https://github.com/mindspore-ai/models/tree/master/research/cv/resnetv2_50_frn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null}],"tasks":[{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[{"method_slug":"1x1-convolution","method_name":"1x1 Convolution"},{"method_slug":"auxiliary-classifier","method_name":"Auxiliary Classifier"},{"method_slug":"average-pooling","method_name":"Average Pooling"},{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"bottleneck-residual-block","method_name":"Bottleneck Residual Block"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"fpn","method_name":"FPN"},{"method_slug":"filter-response-normalization","method_name":"Filter Response Normalization"},{"method_slug":"focal-loss","method_name":"Focal Loss"},{"method_slug":"global-average-pooling","method_name":"Global Average Pooling"},{"method_slug":"inception-v3","method_name":"Inception-v3"},{"method_slug":"inception-v3-module","method_name":"Inception-v3 Module"},{"method_slug":"kaiming-initialization","method_name":"Kaiming Initialization"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"linear-warmup-with-cosine-annealing","method_name":"Linear Warmup With Cosine Annealing"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"rmsprop","method_name":"RMSProp"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"residual-block","method_name":"Residual Block"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"retinanet","method_name":"RetinaNet"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[{"slug":"filter-response-normalization","name":"Filter Response Normalization","full_name":"Filter Response Normalization"}],"results":[{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"InceptionV3 (FRN layer)","rank_in_archive_order":796,"of":1060,"metrics":{"Top 1 Accuracy":"78.95%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ResnetV2 50 (FRN layer)","rank_in_archive_order":878,"of":1060,"metrics":{"Top 1 Accuracy":"77.21%"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1911.09737","atlas_url":"https://app.syntology.ai/?focus=1911.09737","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"1911.09737"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/EmptySamurai/pytorch-reconet","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/mindspore-ai/models/tree/master/research/cv/resnetv2_50_frn","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/MindSpore-paper-code-3/code5/tree/main/resnetv2_50_frn","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/CarloLepelaars/filter_response_normalization_keras","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/gakkiri/Filter-Response-Normalization","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/kobiso/FilterResponseNorm-MXNet","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/2023-MindSpore-1/ms-code-216/tree/main/resnetv2_50_frn","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ensta-u2is/torch-uncertainty","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/gupta-abhay/pytorch-frn","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/2023-MindSpore-4/Code11/tree/main/resnetv2_50_frn","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/tattaka/Filter-Response-Normalization-PyTorch","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ppwwyyxx/FRN-on-common-ImageNet-baseline","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/amirbar/FilterResponseNormalization","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/SdahlSean/FilterResponseNormalizationTensorflow","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/philipperemy/keras-frn","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/2023-MindSpore-4/Code14/tree/main/resnetv2_50_frn","reach":null}],"summary":{"ran_draft_wrong":1,"ran_fixture":1},"by_repo_kind":{"listed":{"samples":2,"ran":2,"repositories":2}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"c9a5f3feedf35dc9","entry":"convert_model","repo":"tattaka/Filter-Response-Normalization-PyTorch","repo_kind":"listed","path":"filter_response_normalization/filter_response_normalize.py","file_url":"https://github.com/tattaka/Filter-Response-Normalization-PyTorch/blob/HEAD/filter_response_normalization/filter_response_normalize.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"c9a5f3feedf35dc9"}},{"code_sha256_prefix":"d799bca8504332a9","entry":"frn_layer_paper","repo":"philipperemy/keras-frn","repo_kind":"listed","path":"frn/frn.py","file_url":"https://github.com/philipperemy/keras-frn/blob/HEAD/frn/frn.py","link_basis":"first_harvest_node","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"d799bca8504332a9"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}