{"url":"/sota/fine-grained-image-classification-on-oxford","task":{"name":"Fine-Grained Image Classification","url":"/task/fine-grained-image-classification","note":null},"dataset":{"name":"Oxford 102 Flowers","url":"/dataset/oxford-102-flower"},"category":"Computer Vision","categories":["Computer Vision"],"category_note":null,"description":"**Fine-Grained Image Classification** is a task in computer vision where the goal is to classify images into subcategories within a larger category. For example, classifying different species of birds or different types of flowers. This task is considered to be fine-grained because it requires the model to distinguish between subtle differences in visual appearance and patterns, making it more challenging than regular image classification tasks.\r\n\r\n<span style=\"color:grey; opacity: 0.6\">( Image credit: [Looking for the Devil in the Details](https://arxiv.org/pdf/1903.06150v2.pdf) )</span>","description_from":"task","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","rank":"the archive's row order at snapshot; not re-ranked","rows_end_at":"2025-07-28","rows_withheld_as_spam":0,"metric_values":"the archive's strings, untouched"},"metrics":["Accuracy","Top-1 Error Rate","FLOPS","PARAMS","Top 1 Accuracy"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Accuracy":"higher","Top-1 Error Rate":null,"FLOPS":"lower","PARAMS":"lower","Top 1 Accuracy":"higher"}},"counts":{"rows":25,"rows_with_code":24,"rows_with_paper_page":25,"rows_dated":25,"rows_using_additional_data":13},"rows":[{"rank_in_archive_order":1,"model":"IELT","metrics":{"Accuracy":"99.64%"},"uses_additional_data":false,"paper_date":"2023-02-13","paper":"/paper/fine-grained-visual-classification-via-2","paper_url":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10042971","paper_title":"Fine-Grained Visual Classification via Internal Ensemble Learning Transformer","code":"https://github.com/mobulan/ielt","n_code_links":1,"syntology":null},{"rank_in_archive_order":2,"model":"BiT-L (ResNet)","metrics":{"Accuracy":"99.63%","Top-1 Error Rate":"0.37"},"uses_additional_data":true,"paper_date":"2019-12-24","paper":"/paper/large-scale-learning-of-general-visual","paper_url":"https://arxiv.org/abs/1912.11370v3","paper_title":"Big Transfer (BiT): General Visual Representation Learning","code":"https://github.com/google-research/big_transfer","n_code_links":9,"syntology":{"n_ran":3,"n_unverified":7,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":3,"model":"µ2Net (ViT-L/16)","metrics":{"Accuracy":"99.61%"},"uses_additional_data":false,"paper_date":"2022-05-25","paper":"/paper/an-evolutionary-approach-to-dynamic","paper_url":"https://arxiv.org/abs/2205.12755v6","paper_title":"An Evolutionary Approach to Dynamic Introduction of Tasks in Large-scale Multitask Learning Systems","code":"https://github.com/google-research/google-research/tree/master/muNet","n_code_links":1,"syntology":null},{"rank_in_archive_order":4,"model":"BiT-M (ResNet)","metrics":{"Accuracy":"99.30%","Top-1 Error Rate":"0.70"},"uses_additional_data":true,"paper_date":"2019-12-24","paper":"/paper/large-scale-learning-of-general-visual","paper_url":"https://arxiv.org/abs/1912.11370v3","paper_title":"Big Transfer (BiT): General Visual Representation Learning","code":"https://github.com/google-research/big_transfer","n_code_links":9,"syntology":{"n_ran":3,"n_unverified":7,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":5,"model":"Wide-ResNet-101 (Spinal FC)","metrics":{"Accuracy":"99.30%"},"uses_additional_data":true,"paper_date":"2020-07-07","paper":"/paper/spinalnet-deep-neural-network-with-gradual-1","paper_url":"https://arxiv.org/abs/2007.03347v3","paper_title":"SpinalNet: Deep Neural Network with Gradual Input","code":"https://github.com/dipuk0506/SpinalNet","n_code_links":3,"syntology":null},{"rank_in_archive_order":6,"model":"TResNet-L","metrics":{"Accuracy":"99.1%"},"uses_additional_data":true,"paper_date":"2020-03-30","paper":"/paper/tresnet-high-performance-gpu-dedicated","paper_url":"https://arxiv.org/abs/2003.13630v3","paper_title":"TResNet: High Performance GPU-Dedicated Architecture","code":"https://github.com/rwightman/pytorch-image-models","n_code_links":3,"syntology":null},{"rank_in_archive_order":7,"model":"Grafit (RegNet-8GF)","metrics":{"Accuracy":"99.1%"},"uses_additional_data":false,"paper_date":"2020-11-25","paper":"/paper/grafit-learning-fine-grained-image","paper_url":"https://arxiv.org/abs/2011.12982v1","paper_title":"Grafit: Learning fine-grained image representations with coarse labels","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":8,"model":"TNT-B","metrics":{"Accuracy":"99.0%","PARAMS":"65.6M"},"uses_additional_data":false,"paper_date":"2021-02-27","paper":"/paper/transformer-in-transformer","paper_url":"https://arxiv.org/abs/2103.00112v3","paper_title":"Transformer in Transformer","code":"https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/tnt.py","n_code_links":12,"syntology":{"n_ran":16,"n_unverified":8,"n_samples":24,"n_pointer_only_licence":5}},{"rank_in_archive_order":9,"model":"Assemble-ResNet","metrics":{"Accuracy":"98.9%"},"uses_additional_data":true,"paper_date":"2020-01-17","paper":"/paper/compounding-the-performance-improvements-of","paper_url":"https://arxiv.org/abs/2001.06268v2","paper_title":"Compounding the Performance Improvements of Assembled Techniques in a Convolutional Neural Network","code":"https://github.com/clovaai/assembled-cnn","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":10,"n_samples":11,"n_pointer_only_licence":0}},{"rank_in_archive_order":10,"model":"DeiT-B","metrics":{"Accuracy":"98.8%","PARAMS":"86M"},"uses_additional_data":false,"paper_date":"2020-12-23","paper":"/paper/training-data-efficient-image-transformers","paper_url":"https://arxiv.org/abs/2012.12877v2","paper_title":"Training data-efficient image transformers & distillation through attention","code":"https://github.com/huggingface/transformers","n_code_links":40,"syntology":{"n_ran":12,"n_unverified":7,"n_samples":19,"n_pointer_only_licence":3}},{"rank_in_archive_order":11,"model":"DenseNet-201(Spinal FC)","metrics":{"Accuracy":"98.36"},"uses_additional_data":false,"paper_date":"2021-10-14","paper":"/paper/a-comprehensive-study-on-torchvision-pre","paper_url":"https://arxiv.org/abs/2110.07097v1","paper_title":"A Comprehensive Study on Torchvision Pre-trained Models for Fine-grained Inter-species Classification","code":"https://github.com/dipuk0506/SpinalNet","n_code_links":1,"syntology":null},{"rank_in_archive_order":12,"model":"NAT-M4","metrics":{"Accuracy":"98.3%","FLOPS":"400M","PARAMS":"4.2M"},"uses_additional_data":true,"paper_date":"2020-05-12","paper":"/paper/neural-architecture-transfer","paper_url":"https://arxiv.org/abs/2005.05859v2","paper_title":"Neural Architecture Transfer","code":"https://github.com/human-analysis/neural-architecture-transfer","n_code_links":2,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":13,"model":"DenseNet-201","metrics":{"Accuracy":"98.29"},"uses_additional_data":false,"paper_date":"2021-10-14","paper":"/paper/a-comprehensive-study-on-torchvision-pre","paper_url":"https://arxiv.org/abs/2110.07097v1","paper_title":"A Comprehensive Study on Torchvision Pre-trained Models for Fine-grained Inter-species Classification","code":"https://github.com/dipuk0506/SpinalNet","n_code_links":1,"syntology":null},{"rank_in_archive_order":14,"model":"NAT-M3","metrics":{"Accuracy":"98.1%","FLOPS":"250M","PARAMS":"3.7M"},"uses_additional_data":true,"paper_date":"2020-05-12","paper":"/paper/neural-architecture-transfer","paper_url":"https://arxiv.org/abs/2005.05859v2","paper_title":"Neural Architecture Transfer","code":"https://github.com/human-analysis/neural-architecture-transfer","n_code_links":2,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":15,"model":"ResNet50 (A1)","metrics":{"Accuracy":"97.9%","FLOPS":"4.1","PARAMS":"24M"},"uses_additional_data":false,"paper_date":"2021-10-01","paper":"/paper/resnet-strikes-back-an-improved-training","paper_url":"https://arxiv.org/abs/2110.00476v1","paper_title":"ResNet strikes back: An improved training procedure in timm","code":"https://github.com/rwightman/pytorch-image-models","n_code_links":14,"syntology":{"n_ran":0,"n_unverified":3,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":16,"model":"SR-GNN","metrics":{"Accuracy":"97.9%","FLOPS":"9.8","PARAMS":"30.9"},"uses_additional_data":false,"paper_date":"2022-09-05","paper":"/paper/sr-gnn-spatial-relation-aware-graph-neural","paper_url":"https://arxiv.org/abs/2209.02109v1","paper_title":"SR-GNN: Spatial Relation-aware Graph Neural Network for Fine-Grained Image Categorization","code":"https://github.com/ardhendubehera/sr-gnn","n_code_links":1,"syntology":null},{"rank_in_archive_order":17,"model":"ResMLP-24","metrics":{"Accuracy":"97.9%"},"uses_additional_data":true,"paper_date":"2021-05-07","paper":"/paper/resmlp-feedforward-networks-for-image","paper_url":"https://arxiv.org/abs/2105.03404v2","paper_title":"ResMLP: Feedforward networks for image classification with data-efficient training","code":"https://github.com/rwightman/pytorch-image-models","n_code_links":19,"syntology":{"n_ran":2,"n_unverified":5,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":18,"model":"NAT-M2","metrics":{"Accuracy":"97.9%","FLOPS":"195M","PARAMS":"3.4M"},"uses_additional_data":true,"paper_date":"2020-05-12","paper":"/paper/neural-architecture-transfer","paper_url":"https://arxiv.org/abs/2005.05859v2","paper_title":"Neural Architecture Transfer","code":"https://github.com/human-analysis/neural-architecture-transfer","n_code_links":2,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":19,"model":"ResMLP-12","metrics":{"Accuracy":"97.4%"},"uses_additional_data":true,"paper_date":"2021-05-07","paper":"/paper/resmlp-feedforward-networks-for-image","paper_url":"https://arxiv.org/abs/2105.03404v2","paper_title":"ResMLP: Feedforward networks for image classification with data-efficient training","code":"https://github.com/rwightman/pytorch-image-models","n_code_links":19,"syntology":{"n_ran":2,"n_unverified":5,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":20,"model":"FixInceptionResNet-V2","metrics":{"Accuracy":"95.7%","Top-1 Error Rate":"4.3%"},"uses_additional_data":true,"paper_date":"2019-06-14","paper":"/paper/fixing-the-train-test-resolution-discrepancy","paper_url":"https://arxiv.org/abs/1906.06423v4","paper_title":"Fixing the train-test resolution discrepancy","code":"https://github.com/facebookresearch/FixRes","n_code_links":3,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":21,"model":"AutoAugment","metrics":{"Accuracy":"95.36%","Top-1 Error Rate":"4.64%"},"uses_additional_data":false,"paper_date":"2018-05-24","paper":"/paper/autoaugment-learning-augmentation-policies","paper_url":"http://arxiv.org/abs/1805.09501v3","paper_title":"AutoAugment: Learning Augmentation Policies from Data","code":"https://github.com/tensorflow/models/tree/master/research/autoaugment","n_code_links":33,"syntology":{"n_ran":6,"n_unverified":37,"n_samples":43,"n_pointer_only_licence":2}},{"rank_in_archive_order":22,"model":"PC Bilinear CNN","metrics":{"Accuracy":"93.65%"},"uses_additional_data":false,"paper_date":"2017-05-22","paper":"/paper/pairwise-confusion-for-fine-grained-visual","paper_url":"http://arxiv.org/abs/1705.08016v3","paper_title":"Pairwise Confusion for Fine-Grained Visual Classification","code":"https://github.com/abhimanyudubey/confusion","n_code_links":1,"syntology":null},{"rank_in_archive_order":23,"model":"AutoFormer-S | 384","metrics":{"Top 1 Accuracy":"98.8"},"uses_additional_data":false,"paper_date":"2021-07-01","paper":"/paper/autoformer-searching-transformers-for-visual","paper_url":"https://arxiv.org/abs/2107.00651v1","paper_title":"AutoFormer: Searching Transformers for Visual Recognition","code":"https://github.com/microsoft/AutoML","n_code_links":2,"syntology":null},{"rank_in_archive_order":24,"model":"CCT-14/7x2","metrics":{"FLOPS":"15G","PARAMS":"22.5M"},"uses_additional_data":true,"paper_date":"2021-04-12","paper":"/paper/escaping-the-big-data-paradigm-with-compact","paper_url":"https://arxiv.org/abs/2104.05704v4","paper_title":"Escaping the Big Data Paradigm with Compact Transformers","code":"https://github.com/keras-team/keras-io/blob/master/examples/vision/cct.py","n_code_links":9,"syntology":{"n_ran":3,"n_unverified":3,"n_samples":6,"n_pointer_only_licence":1}},{"rank_in_archive_order":25,"model":"NAT-M1","metrics":{"FLOPS":"152M","PARAMS":"3.3M"},"uses_additional_data":true,"paper_date":"2020-05-12","paper":"/paper/neural-architecture-transfer","paper_url":"https://arxiv.org/abs/2005.05859v2","paper_title":"Neural Architecture Transfer","code":"https://github.com/human-analysis/neural-architecture-transfer","n_code_links":2,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":1}}],"since_archive":{"claim":"Results that newer papers report for their own method, placed here by Syntology. A model pointed at the cell in the paper's own table; the number was read from that cell and checked against this leaderboard's metric, dataset, split and scale; an independent check that saw this leaderboard's other rows and every other leaderboard on the same dataset accepted it. Not reviewed by the paper's authors or by the archive's editors, and not ranked against the archive rows.","extraction_file_present":true,"measurement":{"test_papers":883,"papers_with_output":881,"judged_true":108,"judged":110,"wilson95_lower":0.9361,"measured_on":"2026-09-24","frozen_commit":"0e3de0df94"},"measurement_note":"blind adjudication of accepted entries on a held-out split of archive papers, rules frozen before the test","coverage":{"sentence":"Syntology has checked 6,264 of the 9,581 papers on this site that are newer than the archive; results from the others appear after they are checked.","complete":false,"papers_newer_than_archive":9581,"papers_checked":6264,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":3316},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-24T18:15:14+00:00","claim":"Per row: N of M harvested code samples from that row's paper executed on a synthesized fixture; the other M-N are unverified. Not a reproduction of the row's number; not a correctness claim. n_pointer_only_licence counts samples the site points at rather than redistributes (a licence axis, independent of ran/unverified).","rows_with_graph_line":15,"rows_with_any_sample_ran":9,"distinct_papers_with_graph_line":10,"distinct_papers_with_any_sample_ran":7,"samples_over_distinct_papers":{"n_ran":43,"n_unverified":83,"n_samples":126,"n_pointer_only_licence":14,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":48,"n_unverified":98,"n_samples":146,"n_pointer_only_licence":17,"note":"row-weighted: a paper behind several rows is counted once per row; inflated relative to samples_over_distinct_papers by design, kept for readers summing the per-row syntology blocks"}}}