Imagenet Subset of Ten Easily Classified Image Categories

Imagenette is a subset of the ImageNet dataset containing 10 easily classified image categories, including tench, English springer, and cassette player. It was created by author frgfm and last updated in December 2022.

Use Cases

• Train an image classifier on the 10 distinct categories like 'golf ball' and 'parachute' for educational or benchmarking purposes.
• Fine-tune a pre-trained vision model on the 'church' and 'French horn' classes to evaluate transfer learning performance.
• Use the 'garbage truck' and 'gas pump' image categories for testing lightweight model architectures on a manageable class set.

Strengths

• Derived from the well-known ImageNet dataset, providing a standardized benchmark foundation.
• Contains 10 distinct and easily classified categories, simplifying initial model training and evaluation.
• Last updated in December 2022, indicating recent maintenance.

Limitations

• Contains only 10 classes, which limits its utility for training models requiring fine-grained or numerous categories.
• The dataset is a curated subset, which may not represent the full complexity or diversity of the original ImageNet data.
• Specific details on dataset size, image resolution, and class balance are not provided in the input.

Provenance

Source

Subset of the ImageNet dataset.

Collection Method

Crowdsourced annotations, selected as 10 easily classified classes from ImageNet.

Freshness

Last updated on 2022-12-11.

Geography

Region is indicated as 'us' in tags, but specific spatial coverage is unknown.