MixCount: Synthetic Images for Mixed-Object, Open-Vocabulary Counting

A large-scale synthetic dataset for mixed-object, open-vocabulary counting, a setting that reportedly breaks current counting models. The dataset was created by researchers from EPFL, the University of Oxford, and Northwestern University. It was last updated on the Hugging Face platform in May 2026.

Use Cases

• Training object counting models for industrial inspection based on the described focus on sorting tasks.
• Benchmarking model performance on open-vocabulary counting scenarios as mentioned in the description.
• Developing fine-grained counting algorithms using the synthetic images and multimodal prompts indicated by platform tags.

Strengths

• Dataset is described as large-scale, suggesting a substantial number of samples.
• Created by a collaboration of researchers from EPFL, the University of Oxford, and Northwestern University.
• Focuses on a specific, challenging problem (mixed-object, open-vocabulary counting) for industrial applications.

Limitations

• Row count, file formats, and column-level documentation are unknown, limiting suitability assessment.
• Data is synthetic, which may not fully capture the complexity and noise of real-world industrial environments.
• License information is unavailable, which could restrict commercial use.

Provenance

Source

Hugging Face, uploaded by author 'CorentinDumery'.

Collection Method

Automatically generated via a synthetic pipeline, as mentioned in the description.

Freshness

Last updated 2026-05-19 15:07:06.