# Object Detection on Open Images unseen classes (non-overlapping) Source: [Object Detection on Open Images unseen classes (non-overlapping)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:23:36.939Z Evaluation of Object Detection performance on the Open Images dataset, specifically focusing on classes that are non-overlapping (unseen during training), measured using Average Recall (AR@1000) across different object sizes. 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 | Object Detection | | Dataset (including any stated split) | Open Images unseen classes (non-overlapping) | | Metric | AR@1000 (Overall) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 11 of 11 | | 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 - [AR@1000 (Overall)](): Higher is better; unit: PERCENT. - [AR@1000 (Small)](): Higher is better; unit: PERCENT. - [AR@1000 (Medium)](): Higher is better; unit: PERCENT. - [AR@1000 (Large)](): Higher is better; unit: PERCENT. ## Recorded results | Method | AR@1000 (Overall) | Result date | Report paper | | --- | --- | --- | --- | | SSD-ag-ad | 21 | Not recorded | 68fa68da1e209131cea1eb24 | | SSD-ag | 20.6 | Not recorded | 68fa68da1e209131cea1eb24 | | SSD-aw | 19.9 | Not recorded | 68fa68da1e209131cea1eb24 | | SSD-ft-ag-ad | 19.4 | Not recorded | 68fa68da1e209131cea1eb24 | | FRCNN-ag-ad | 19.2 | Not recorded | 68fa68da1e209131cea1eb24 | | SSD-ft-ag | 18.8 | Not recorded | 68fa68da1e209131cea1eb24 | | FRCNN-ft-ag | 18.4 | Not recorded | 68fa68da1e209131cea1eb24 | | FRCNN-ag | 18.2 | Not recorded | 68fa68da1e209131cea1eb24 | | FRCNN-ft-ag-ad | 18.1 | Not recorded | 68fa68da1e209131cea1eb24 | | FRCNN-aw | 17 | Not recorded | 68fa68da1e209131cea1eb24 | | FRCNN-aw-prop | 15.3 | Not recorded | 68fa68da1e209131cea1eb24 | ### SSD-ag-ad AR@1000 (Overall): 21. Result ID: 698d89898b517eb4332238f2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Architecture":"SSD","Class Awareness":"Agnostic","Training Protocol":"Scratch","Training Method":"Adversarial"}. This record does not assert independent reproduction. ### SSD-ag AR@1000 (Overall): 20.6. Result ID: 698d89898b517eb4332238f0. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Base Architecture":"SSD","Class Awareness":"Agnostic","Training Protocol":"Scratch"}. This record does not assert independent reproduction. ### SSD-aw AR@1000 (Overall): 19.9. Result ID: 698d89898b517eb4332238ee. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Base Architecture":"SSD","Class Awareness":"Aware"}. This record does not assert independent reproduction. ### SSD-ft-ag-ad AR@1000 (Overall): 19.4. Result ID: 698d89898b517eb4332238f1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Architecture":"SSD","Class Awareness":"Agnostic","Training Protocol":"Fine-tuned from class-aware baseline","Training Method":"Adversarial"}. This record does not assert independent reproduction. ### FRCNN-ag-ad AR@1000 (Overall): 19.2. Result ID: 698d89898b517eb4332238ed. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Architecture":"FRCNN","Class Awareness":"Agnostic","Training Protocol":"Scratch","Training Method":"Adversarial"}. This record does not assert independent reproduction. ### SSD-ft-ag AR@1000 (Overall): 18.8. Result ID: 698d89898b517eb4332238ef. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Base Architecture":"SSD","Class Awareness":"Agnostic","Training Protocol":"Fine-tuned from class-aware baseline"}. This record does not assert independent reproduction. ### FRCNN-ft-ag AR@1000 (Overall): 18.4. Result ID: 698d89898b517eb4332238ea. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Base Architecture":"FRCNN","Class Awareness":"Agnostic","Training Protocol":"Fine-tuned from class-aware baseline"}. This record does not assert independent reproduction. ### FRCNN-ag AR@1000 (Overall): 18.2. Result ID: 698d89898b517eb4332238eb. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Base Architecture":"FRCNN","Class Awareness":"Agnostic","Training Protocol":"Scratch"}. This record does not assert independent reproduction. ### FRCNN-ft-ag-ad AR@1000 (Overall): 18.1. Result ID: 698d89898b517eb4332238ec. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Architecture":"FRCNN","Class Awareness":"Agnostic","Training Protocol":"Fine-tuned from class-aware baseline","Training Method":"Adversarial"}. This record does not assert independent reproduction. ### FRCNN-aw AR@1000 (Overall): 17. Result ID: 698d89898b517eb4332238e9. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Base Architecture":"FRCNN","Class Awareness":"Aware"}. This record does not assert independent reproduction. ### FRCNN-aw-prop AR@1000 (Overall): 15.3. Result ID: 698d89898b517eb4332238e8. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Base Architecture":"FRCNN","Class Awareness":"Aware","Evaluation Stage":"First-stage Region Proposal Network"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.