Micro Facial Expressions: 10K-100K Merged Grayscale and Color Images

Aggregating between 10,000 and 100,000 facial images, this collection for emotion recognition was compiled by LaurenGurgiolo in 2025. It merges the FER-2013 grayscale dataset with a color Micro-Expression Image Dataset to support affective computing tasks.

Use Cases

• Training image classification models to categorize facial states using the provided emotion labels
• Evaluating model generalization across different image modalities by comparing performance on grayscale FER-2013 vs color micro-expression samples
• Developing affective computing pipelines specifically targeting the detection of subtle micro-expressions

Strengths

• 10,000 to 100,000 image records
• Combines grayscale and color image modalities
• Includes specialized micro-expression labels

Limitations

• Inconsistent image formats (grayscale vs color) due to merged sources
• Source licenses are inherited and require individual verification

Provenance

Source

FER-2013 and Micro-Expression Image Dataset

Collection Method

Aggregated from existing facial expression datasets

Freshness

Last updated December 2025.