Snack Food Images with Bounding Box Annotations

Images of 20 different snack foods, sourced from the Google Open Images dataset release 2017_11. It includes CSV files with bounding box annotations for object detection, though not all images are annotated and some have multiple annotations. The dataset was created by Matthijs to accompany a machine learning book.

Use Cases

• Train an object detection model to locate and classify snack foods using the provided bounding box annotations.
• Analyze the distribution of snack types and the frequency of multiple annotations per image.
• Fine-tune a pre-trained vision model on the 20 snack categories for image classification tasks.

Strengths

• Contains 20 distinct snack food categories for model training.
• Annotations include bounding boxes, enabling object detection tasks.
• Sourced from the established Google Open Images dataset.

Limitations

• Not all images have bounding box annotations, reducing usable samples for detection.
• The underlying image source (Google Open Images 2017_11) may be temporally stale.
• Dataset size and specific row/column counts are not provided, limiting assessment of scale.

Provenance

Source

Google Open Images dataset, release 2017_11.

Collection Method

Images were taken and annotated from the specified Open Images release.

Freshness

Last updated on 2022-04-12.

Geography

Region tag indicates 'us', suggesting a potential US focus, but not explicitly confirmed.