# Few-shot Character Recognition on Omniglot (novel classes) Source: [Few-shot Character Recognition on Omniglot (novel classes)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:27:35.542Z Evaluation of Few-shot Character Recognition performance on the novel classes split of the Omniglot dataset, typically measured using 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 | Few-shot Character Recognition | | Dataset (including any stated split) | Omniglot (novel classes) | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 12 of 12 | | 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 | | --- | --- | --- | --- | | MAML | 99.53 | Not recorded | 68fa66e41e209131cea12e8a | | Baseline++ | 99.38 | Not recorded | 68fa66e41e209131cea12e8a | | MatchingNet | 99.37 | Not recorded | 68fa66e41e209131cea12e8a | | RelationNet | 99.3 | Not recorded | 68fa66e41e209131cea12e8a | | ProtoNet | 99.15 | Not recorded | 68fa66e41e209131cea12e8a | | Baseline | 99.12 | Not recorded | 68fa66e41e209131cea12e8a | | MAML | 98.57 | Not recorded | 68fa66e41e209131cea12e8a | | ProtoNet | 98.01 | Not recorded | 68fa66e41e209131cea12e8a | | MatchingNet | 97.78 | Not recorded | 68fa66e41e209131cea12e8a | | RelationNet | 97.22 | Not recorded | 68fa66e41e209131cea12e8a | | Baseline++ | 95.41 | Not recorded | 68fa66e41e209131cea12e8a | | Baseline | 94.89 | Not recorded | 68fa66e41e209131cea12e8a | ### MAML Accuracy: 99.53. Result ID: 698d0e838b517eb4331eaa34. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"N-shot":"5-shot","Backbone":"Conv-4","Data Augmentation":"None","N-way":"5-way"}. This record does not assert independent reproduction. ### Baseline++ Accuracy: 99.38. Result ID: 698d0e838b517eb4331eaa31. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"N-shot":"5-shot","Backbone":"Conv-4","Data Augmentation":"None","N-way":"5-way","Training Epochs":5}. This record does not assert independent reproduction. ### MatchingNet Accuracy: 99.37. Result ID: 698d0e838b517eb4331eaa32. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"N-shot":"5-shot","Backbone":"Conv-4","Data Augmentation":"None","N-way":"5-way"}. This record does not assert independent reproduction. ### RelationNet Accuracy: 99.3. Result ID: 698d0e838b517eb4331eaa35. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"N-shot":"5-shot","Backbone":"Conv-4","Data Augmentation":"None","N-way":"5-way"}. This record does not assert independent reproduction. ### ProtoNet Accuracy: 99.15. Result ID: 698d0e838b517eb4331eaa33. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"N-shot":"5-shot","Backbone":"Conv-4","Data Augmentation":"None","N-way":"5-way"}. This record does not assert independent reproduction. ### Baseline Accuracy: 99.12. Result ID: 698d0e838b517eb4331eaa30. [Reporting paper]() Recorded evaluation configuration: {"N-shot":"5-shot","Backbone":"Conv-4","Data Augmentation":"None","N-way":"5-way","Training Epochs":5}. This record does not assert independent reproduction. ### MAML Accuracy: 98.57. Result ID: 698d0e838b517eb4331eaa2e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"N-shot":"1-shot","Backbone":"Conv-4","Data Augmentation":"None","N-way":"5-way"}. This record does not assert independent reproduction. ### ProtoNet Accuracy: 98.01. Result ID: 698d0e838b517eb4331eaa2d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"N-shot":"1-shot","Backbone":"Conv-4","Data Augmentation":"None","N-way":"5-way"}. This record does not assert independent reproduction. ### MatchingNet Accuracy: 97.78. Result ID: 698d0e838b517eb4331eaa2c. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"N-shot":"1-shot","Backbone":"Conv-4","Data Augmentation":"None","N-way":"5-way"}. This record does not assert independent reproduction. ### RelationNet Accuracy: 97.22. Result ID: 698d0e838b517eb4331eaa2f. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"N-shot":"1-shot","Backbone":"Conv-4","Data Augmentation":"None","N-way":"5-way"}. This record does not assert independent reproduction. ### Baseline++ Accuracy: 95.41. Result ID: 698d0e838b517eb4331eaa2b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"N-shot":"1-shot","Backbone":"Conv-4","Data Augmentation":"None","N-way":"5-way","Training Epochs":5}. This record does not assert independent reproduction. ### Baseline Accuracy: 94.89. Result ID: 698d0e838b517eb4331eaa2a. [Reporting paper]() Recorded evaluation configuration: {"N-shot":"1-shot","Backbone":"Conv-4","Data Augmentation":"None","N-way":"5-way","Training Epochs":5}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.