Datasets:

mushroomfleet
/

reynolds-oil-style

Modalities:

Image

Text

Formats:

imagefolder

Size:

< 1K

Libraries:

Datasets

Dataset card Data Studio Files Files and versions

xet

Community

Dataset Viewer

Auto-converted to Parquet

Split (1)

train · 148 rows

Search is not available for this dataset

image imagewidth (px) 400 1.92k

End of preview. Expand in Data Studio

YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/datasets-cards)

Reynolds Oil Style Dataset

This repository contains the Reynolds Oil Style dataset with multiple captioning methodologies for training and evaluation purposes.

Repository Structure

Folders

Each folder contains the images and captions for different dataset versions:

reynolds-oil-style/ - Base dataset (39 files)
reynolds-oil-style-b/ - "[Trigger Classifier] Prefix" method version (37 files)
reynolds-oil-style-nocap/ - No captions version for recaptioning tests (19 files)
reynoldsv0d-recontext/ - "In-context [Trigger Classifier]" method version (37 files)

Zip Files

Pre-packaged versions for direct URL training:

reynolds-oil-style.zip - Base dataset package (3.68 MB)
reynolds-oil-style-b.zip - "[Trigger Classifier] Prefix" method package (3.05 MB)
reynolds-oil-style-nocap.zip - No captions package (3.04 MB)
reynoldsv0d-recontext.zip - "In-context [Trigger Classifier]" method package (3.05 MB)

Dataset Versions

Base Dataset (`reynolds-oil-style`)

The standard version of the Reynolds Oil Style dataset containing original images and captions.

Filename Suffix Explanations

Suffix	Description	Method
`-nocap`	Features no captions	Used for recaptioning tests and evaluation
`-b`	Video captions with "video" → image classifier swap	"[Trigger Classifier] Prefix" method
`v0d-recontext`	In-context trigger classifier captioning	"In-context [Trigger Classifier]" method

Usage

For Local Development

Download the appropriate folder for your use case
Extract and use the images and captions as needed

For Direct URL Training

Use the zip files for streamlined training pipelines
The zip files can be directly referenced by URL in training scripts
Choose the appropriate version based on your captioning methodology requirements

Captioning Methods

"[Trigger Classifier] Prefix" Method (`-b` suffix)

This version modifies video captions by replacing the word "video" with appropriate image classifier terms, creating a prefix-based trigger classification approach.

"In-context [Trigger Classifier]" Method (`v0d-recontext` suffix)

This version uses in-context learning approaches for trigger classification, providing contextual captioning that adapts to the specific image content.