Meta Album Cars Dataset (Micro): 800 Images of 20 Car Models

Meta_Album_CRS_Micro is a subset of the Stanford Cars Dataset, containing 800 square images of cars resized to 128x128 pixels. The dataset includes 20 classes of car models, with images originally collected in 2013 by Jonathan Krause, Michael Stark, Jia Deng, and Li Fei-Fei. This preprocessed version was created by Philip Boser for Meta-Album in March 2022.

Use Cases

• Fine-grained classification of car models based on image features.
• Benchmarking few-shot learning algorithms on a standardized vehicle image dataset.
• Training computer vision models for object recognition in constrained resolution (128x128).
• Studying preprocessing techniques for image datasets where cropping loses information.

Strengths

• Contains 800 images across 20 distinct car model classes.
• All images are preprocessed to a uniform 128x128 pixel resolution.
• Dataset is part of the Meta-Album benchmark collection for standardized evaluation.

Limitations

• Dataset size is limited to 800 images, which may not be sufficient for training large models.
• Image resolution is fixed at 128x128, potentially limiting detail for fine-grained tasks.
• The original data was collected in 2013; newer car models are not represented.

Provenance

Source

Stanford Cars Dataset (https://ai.stanford.edu/~jkrause/cars/car_dataset.html)

Collection Method

Images were preprocessed by creating square versions via border duplication and resizing.

Time Range

Original data collected in 2013.