[SAMPLE] Nexdata | Lip Multimodal Data | 2,000 ID |Lip Sync Data |Audio Image AI Training Data | ...

1. Specifications
Data size : 2,000 id, each person collects the audio and video data from 13 different angles +1 txt document

People distribution : race distribution: Asian, Caucasian, Black, Brown, gender distribution: gender balance, age distribution: people aged 18-60

Collecting environment : indoor natural light scenes, indoor fluorescent lamp scenes

Annotated Imagery Data diversity : including multiple scenes, different ages, different shooting angles

Device : cellphone, the resolution is 1,920*1,080

Collecting angle : audio and video data of front face, 3 angles left side face, 3 angles right side face, looking down, looking up, left side face down, right side face down, left side face up and right side face up all 13 different angles were collected at the same time

Recording content : general field, unlimited content

Language : 10 languages, each video is more than 20 seconds

Data format : the video data format is .mp4, the audio is greater than or equal to 16KHz, 16bit, the frame rate is 25-30 fps

Accuracy rata : the accuracy rate of sentence is more than 95%

2. About Nexdata
Nexdata owns off-the-shelf 200,000 hours of speech recognition data, 800TB of Annotated Imagery Data, about 2 billion pieces of Natural Language Processing (NLP) Data. These ready-to-go Annotated Imagery Data support instant delivery, quickly improve the accuracy of AI models. For more details, please visit us at https://www.nexdata.ai/datasets/computervision?source=Datarade

搜集汇总

背景与挑战

背景概述

该数据集包含2000个ID的唇部多模态数据，每个ID采集13个不同角度的音频和视频，并附带文本文档。数据覆盖多种种族、性别平衡的18-60岁人群，在室内自然光和荧光灯环境下录制，涉及10种语言，视频格式为.mp4，音频质量高，句子准确率超过95%。