MELAB
Database Access
Select a database below. Each database page includes: dataset overview, a blank License Agreement PDF template, and an external application link (for submitting info and uploading the signed agreement).
CASME Database
Spontaneous Micro-ExpressionsA spontaneous micro-expression database. From ~1500 elicited facial movements recorded at 60 fps, 195 micro-expressions were selected and annotated with onset/peak/offset frames, action units (AUs), and emotion labels.
CASME II
200 fps · AU & Emotion LabelsAn improved micro-expression database with higher temporal resolution (200 fps) and facial area resolution (~280×340). 247 micro-expressions were selected from nearly 3000 facial movements, with AU and emotion labels and baseline evaluation.
CAS(ME)²
Macro + Micro in Long VideosA database containing both macro- and micro-expressions in long videos. It provides long videos (Part A) and cropped samples (Part B), with labels based on AUs, self-reported emotion for each facial movement, and the emotion types of eliciting videos.
CAS(ME)3
Large-scale · RGB-D · Multi-modalA large-scale spontaneous ME database with ~80 hours of videos (>8,000,000 frames), including 1,109 micro-expressions and 3,490 macro-expressions. It additionally provides depth modality and a large set of unlabeled videos for unsupervised methods.
CASMEMG
Micro-Expressions + Facial EMGA database with micro-expression videos and facial electromyography (EMG). Videos were captured in 4K (3840×2160) at 30 fps; EMG signals were collected from 7 superficial muscles. It contains 147 macro- and 233 micro-expression samples with keyframes, AUs, and emotion labels.
Chinese Dysarthria Speech Database (CDSD)
133 hours · 44 speakersThe largest Chinese dysarthria dataset to date, featuring 133 hours of recordings from 44 speakers, developed to support dysarthric speech recognition research and benchmarks (CER).