# Image Classification on iNaturalist 2021 (val) Source: [Image Classification on iNaturalist 2021 (val)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:30:08.033Z Evaluation of Image Classification performance on the validation split of the iNaturalist 2021 dataset, typically measured by Accuracy. 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. | Scope | Recorded value | | --- | --- | | Task | Image Classification | | Dataset (including any stated split) | iNaturalist 2021 (val) | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 15 of 15 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | ViT-B w/ AGLU | 85.1 | Not recorded | 68fa75a11e209131cea69111 | | ViT-B | 84.7 | Not recorded | 68fa75a11e209131cea69111 | | SE-ResNet50 w/ APA\* + AGLU | 83.9 | Not recorded | 68fa75a11e209131cea69111 | | LoSA | 83.8 | Not recorded | 68fa720f1e209131cea543e5 | | LoRA | 83.3 | Not recorded | 68fa720f1e209131cea543e5 | | Full fine-tuning | 82.5 | Not recorded | 68fa720f1e209131cea543e5 | | SE-ResNet50 | 82.1 | Not recorded | 68fa75a11e209131cea69111 | | Finetune only attention | 81.9 | Not recorded | 68fa720f1e209131cea543e5 | | Finetune only mlp | 81.7 | Not recorded | 68fa720f1e209131cea543e5 | | Finetune last 2 layers | 79.8 | Not recorded | 68fa720f1e209131cea543e5 | | Finetune last layer | 79.2 | Not recorded | 68fa720f1e209131cea543e5 | | Prompt Tuning | 76.8 | Not recorded | 68fa720f1e209131cea543e5 | | Bitfit | 76.4 | Not recorded | 68fa720f1e209131cea543e5 | | LST | 76 | Not recorded | 68fa720f1e209131cea543e5 | | Linear probing | 70.1 | Not recorded | 68fa720f1e209131cea543e5 | ### ViT-B w/ AGLU Accuracy: 85.1. Result ID: 6993dc1d3a6c31484919095d. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ViT-B","AGLU":"true","Crop Size":"224"}. This record does not assert independent reproduction. ### ViT-B Accuracy: 84.7. Result ID: 6993dc1d3a6c31484919095c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-B","Crop Size":"224"}. This record does not assert independent reproduction. ### SE-ResNet50 w/ APA\* + AGLU Accuracy: 83.9. Result ID: 6993dc1d3a6c31484919095b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ResNet-50","SE":"true","APA":"true","AGLU":"true","Crop Size":"224"}. This record does not assert independent reproduction. ### LoSA Accuracy: 83.8. Result ID: 699203a83a6c3148490c0db1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-g (1B)"}. This record does not assert independent reproduction. ### LoRA Accuracy: 83.3. Result ID: 699203a83a6c3148490c0db0. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ViT-g (1B)"}. This record does not assert independent reproduction. ### Full fine-tuning Accuracy: 82.5. Result ID: 699203a83a6c3148490c0da7. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ViT-g (1B)"}. This record does not assert independent reproduction. ### SE-ResNet50 Accuracy: 82.1. Result ID: 6993dc1d3a6c31484919095a. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ResNet-50","SE":"true","Crop Size":"224"}. This record does not assert independent reproduction. ### Finetune only attention Accuracy: 81.9. Result ID: 699203a83a6c3148490c0daf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-g (1B)"}. This record does not assert independent reproduction. ### Finetune only mlp Accuracy: 81.7. Result ID: 699203a83a6c3148490c0dae. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-g (1B)"}. This record does not assert independent reproduction. ### Finetune last 2 layers Accuracy: 79.8. Result ID: 699203a83a6c3148490c0dac. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"ViT-g (1B)"}. This record does not assert independent reproduction. ### Finetune last layer Accuracy: 79.2. Result ID: 699203a83a6c3148490c0daa. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"ViT-g (1B)"}. This record does not assert independent reproduction. ### Prompt Tuning Accuracy: 76.8. Result ID: 699203a83a6c3148490c0dab. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ViT-g (1B)"}. This record does not assert independent reproduction. ### Bitfit Accuracy: 76.4. Result ID: 699203a83a6c3148490c0dad. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ViT-g (1B)"}. This record does not assert independent reproduction. ### LST Accuracy: 76. Result ID: 699203a83a6c3148490c0da9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-g (1B)"}. This record does not assert independent reproduction. ### Linear probing Accuracy: 70.1. Result ID: 699203a83a6c3148490c0da8. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ViT-g (1B)"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.