CAS-VSR-MOV20

March 13, 2026 · View on GitHub

samples

Introduction

The dataset consists of video clips sourced from publicly available platforms such as YouTube and movies.

Each segment is no longer than 3 minutes and covers a variety of visual and speaking conditions, including diverse lighting environments, different resolutions, and significant variations in pose, as shown above.

The video clips of each speech are provided in the format of visual frames, which are given at a reduced resolution of 96x96 pixels, focusing solely on the lip region.

Dataset Split

Approximately one hour of evaluation data was provided, with half used as CAS-VSR-MOV20-Val for preliminary validation and the other half as CAS-VSR-MOV20-Test for the final validation.

Specifically, CAS-VSR-MOV20 includes a total of 2655 samples, with 1335 samples in the validation set and 1320 samples in the test set.

File Structure and Contents

The dataset is organized into the following structure:

CAS-VSR-MOV20/
├── lip_imgs_96/
│   ├── val
│   │   ├── 0a4bdfb250b1b1a071b1a778486391c2.zip
│   │   ├── 0a9bde947751e803d29ed52e012b00b4.zip
│   │   ├── ...
│   ├── test
├── manifest/
│   ├── mov20_id_test.csv #  file_id for test set
│   ├── mov20_id_val.csv #  file_id for val set

Accessing the Dataset

To access the CAS-VSR-MOV20 dataset, please scan the signed agreement here and send it to lipreading@vipl.ict.ac.cn. Please note that the dataset is only available to universities and research institutes for research purposes only. Note that the agreement should be signed by a full-time staff member (usually your tutor). Sharing the dataset with others is not allowed under the terms of the agreement.

CAS-VSR-W1k (LRW-1000): The first in-the-wild Mandarian lip-reading dataset

CAS-VSR-S68: A sentence-level audio-visual speech dataset for speaker-adaptive lip-reading

CAS-VSR-S101: A wild sentence-level audio-visual speech dataset