LLM-Pack is a small object detection and scene understanding dataset containing 40 images of tabletop grocery scenes. The dataset, created by Yannik019, includes annotated item names and object locations. It was last updated on 2026-05-08.
Use Cases
- Train object detection models based on annotated object locations.
- Evaluate scene understanding systems based on cluttered grocery scenarios.
- Benchmark object counting algorithms based on scenes with 6 to 20 items.
- Develop multimodal reasoning systems based on combined image and text annotations.
Strengths
- Contains 40 annotated images.
- Scenes have varying object counts from 6 to 20 items, providing a range of complexity.
Limitations
- Dataset scale is small with only 40 images.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
Provenance
- Source
- Yannik019
- Freshness
- Last updated 2026-05-08 15:41:58; freshness should be verified.