MEP-3M

A large multimodal ecommerce product benchmark joining product imagery with Chinese text and category labels.

What the resource contains

  • About 3 million image-text product examples
  • Hierarchical product categories
  • 599 fine-grained category labels

Potential uses

  • Train product classification
  • Evaluate image-text alignment
  • Research Chinese multimodal product models

Access & formats

Images / text / labels. External dependencies. Follow the project’s instructions on GitHub; some resources require external files, registration or approval.

License & reuse

Dataset instructions limit use to academic research and prohibit commercial use, even though repository code has an MIT license.

Important limitations

The full release is large. Confirm storage, access and category definitions before planning a training pipeline.

Source & attribution

ChenDelong1999 — original GitHub project ↗

Documentation reviewed 2026-09-24. Review scope and licensing notes. Suggest a correction.

Related resources

Fashion & product vision

Fashion-MNIST

zalandoresearch

A compact fashion-image classification benchmark suitable for learning and quick model comparisons.

Open license stated

IDX / image arrays · Repository / linked release

Fashion & product vision

FEIDEGGER

zalandoresearch

Fashion photographs paired with multiple German descriptions for multimodal retrieval research.

Review required

Images / text annotations · Repository / linked release

Fashion & product vision

DeepFashion2

switchablenorms

Clothing images with detailed annotations for detection, segmentation and consumer-to-shop retrieval.

Review required

Images / JSON annotations · External dependencies