# Audio-Visual Speech Recognition on MISP 2021 (test) Source: [Audio-Visual Speech Recognition on MISP 2021 (test)]() Collection index: [llms.txt]() Data updated: 2026-06-05T08:53:06.096Z Evaluation of Audio-Visual Speech Recognition performance on the test split of the MISP 2021 dataset, measured using Character Error Rate (CER) and related metrics. 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 | Audio-Visual Speech Recognition | | Dataset (including any stated split) | MISP 2021 (test) | | Metric | CER | | Unit | SCALAR | | Direction | Lower is better | | Results loaded | 13 of 13 | | 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 - [CER](): Lower is better; unit: SCALAR. - [cpCER](): Lower is better; unit: SCALAR. ## Recorded results | Method | CER | Result date | Report paper | | --- | --- | --- | --- | | M2S-AVROVER | 0.1882 | Not recorded | 6a22589d86173ad9c1468ba7 | | MDA-KD | 0.2153 | Not recorded | 68fa72791e209131cea5698d | | M2S-AV | 0.2195 | Not recorded | 6a22589d86173ad9c1468ba7 | | ModalBiasAVSR | 0.2213 | Not recorded | 6a22589d86173ad9c1468ba7 | | USTC | 0.2458 | Not recorded | 68fa72791e209131cea5698d | | USTC | 0.2458 | Not recorded | 6a22589d86173ad9c1468ba7 | | NIO | 0.2507 | Not recorded | 68fa72791e209131cea5698d | | NIO | 0.2507 | Not recorded | 6a22589d86173ad9c1468ba7 | | M2S-A | 0.2508 | Not recorded | 6a22589d86173ad9c1468ba7 | | Whisper-Flamingo | 0.2608 | Not recorded | 6a22589d86173ad9c1468ba7 | | SJTU | 0.3402 | Not recorded | 68fa72791e209131cea5698d | | SJTU | 0.3402 | Not recorded | 6a22589d86173ad9c1468ba7 | | M2S-V | 0.7635 | Not recorded | 6a22589d86173ad9c1468ba7 | ### M2S-AVROVER CER: 0.1882. Result ID: 6a228e71fbc373592c20c65f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Year":"2026","Training Data (hrs) - Audio":"600†","Training Data (hrs) - Video":"LRS3+Vox2(En)","Backbone":"Transformer"}. This record does not assert independent reproduction. ### MDA-KD CER: 0.2153. Result ID: 69923b403a6c3148490d7dd5. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Training Data (Audio)":"1000 hours","Training Data (Video)":"w/o extra data","Backbone":"Conformer","Objective Function":"ED + InterCTC"}. This record does not assert independent reproduction. ### M2S-AV CER: 0.2195. Result ID: 6a228e71fbc373592c20c65e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Year":"2026","Training Data (hrs) - Audio":"600†","Training Data (hrs) - Video":"LRS3+Vox2(En)","Backbone":"Transformer"}. This record does not assert independent reproduction. ### ModalBiasAVSR CER: 0.2213. Result ID: 6a228e71fbc373592c20c65a. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Year":"2024","Training Data (hrs) - Audio":"1000","Training Data (hrs) - Video":"w/o extra","Backbone":"Conformer"}. This record does not assert independent reproduction. ### USTC CER: 0.2458. Result ID: 69923b403a6c3148490d7dd4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training Data (Audio)":"500 hours","Training Data (Video)":"w/o extra data","Backbone":"Conformer","Objective Function":"ED"}. This record does not assert independent reproduction. ### USTC CER: 0.2458. Result ID: 6a228e71fbc373592c20c659. [Reporting paper]() Recorded evaluation configuration: {"Year":"2023","Training Data (hrs) - Audio":"500","Training Data (hrs) - Video":"w/o extra","Backbone":"Conformer"}. This record does not assert independent reproduction. ### NIO CER: 0.2507. Result ID: 69923b403a6c3148490d7dd3. [Reporting paper]() Recorded evaluation configuration: {"Training Data (Audio)":"3300 hours","Training Data (Video)":"LRW-1000","Backbone":"Transformer","Objective Function":"ED"}. This record does not assert independent reproduction. ### NIO CER: 0.2507. Result ID: 6a228e71fbc373592c20c658. [Reporting paper]() Recorded evaluation configuration: {"Year":"2022","Training Data (hrs) - Audio":"3300","Training Data (hrs) - Video":"LRW-1000","Backbone":"Transformer"}. This record does not assert independent reproduction. ### M2S-A CER: 0.2508. Result ID: 6a228e71fbc373592c20c65c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Year":"2026","Training Data (hrs) - Audio":"600†","Training Data (hrs) - Video":"–","Backbone":"Transformer"}. This record does not assert independent reproduction. ### Whisper-Flamingo CER: 0.2608. Result ID: 6a228e71fbc373592c20c65b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Year":"2025","Training Data (hrs) - Audio":"600†","Training Data (hrs) - Video":"LRS3+Vox2(En)","Backbone":"Transformer"}. This record does not assert independent reproduction. ### SJTU CER: 0.3402. Result ID: 69923b403a6c3148490d7dd2. [Reporting paper]() Recorded evaluation configuration: {"Training Data (Audio)":"300 hours","Training Data (Video)":"LRW-1000","Backbone":"Conformer","Objective Function":"ED + SE"}. This record does not assert independent reproduction. ### SJTU CER: 0.3402. Result ID: 6a228e71fbc373592c20c657. [Reporting paper]() Recorded evaluation configuration: {"Year":"2022","Training Data (hrs) - Audio":"300","Training Data (hrs) - Video":"LRW-1000 \[90\]","Backbone":"Conformer"}. This record does not assert independent reproduction. ### M2S-V CER: 0.7635. Result ID: 6a228e71fbc373592c20c65d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Year":"2026","Training Data (hrs) - Audio":"–","Training Data (hrs) - Video":"LRS3+Vox2(En)","Backbone":"Transformer"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.