Fine-Grained Vehicle Detection Dataset with Corolla and BMW 3-Series Annotations

1410 unique dashcam frames from the BDD100K corpus are annotated for fine-grained vehicle make identification. The dataset contains 1524 annotations, specifically identifying 558 Toyota Corolla sedans and 966 BMW 3-Series sedans. It was created by author 'arrmlet' and last updated on April 28, 2026.

Use Cases

• Train fine-grained vehicle classification models based on the specific make annotations for Toyota Corolla and BMW 3-Series.
• Benchmark object detection performance on real-world dashcam imagery based on the BDD100K-derived bounding boxes.
• Study domain adaptation for vehicle recognition models using the real-world, unconstrained dashcam source imagery.

Strengths

• Contains 1524 fine-grained annotations across 1410 unique images.
• Focuses on two specific, common vehicle makes (Toyota Corolla and BMW 3-Series) for precise classification tasks.
• Derived from the established BDD100K dashcam corpus, providing real-world context.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic or temporal bias inherent to the BDD100K source corpus.

Provenance

Source

BDD100K dashcam corpus (Berkeley DeepDrive)

Collection Method

Fine-grained make annotations added on top of BDD100K's existing car bounding boxes.

Freshness

Last updated 2026-04-28 08:03:03; freshness should be verified.