license: mit
dataset_info:
features:
- name: filename
dtype: string
- name: image
dtype: image
Not a Hot-dog
(The name of this dataset is a reference to the show Silicon Valley.)
Dataset Details
Dataset Description
This dataset is a small collection of user-submitted images that contain objects that are not hot dogs, but might be perceptibly shaped like one.
- Curated by: Public users
- Language(s) (NLP): None (no captions included)
- License: MIT
Dataset Sources
The images provided were submitted by random internet users on Reddit.
Uses
This dataset may be used to train safety checking neural networks, or low-rank adaptation networks that might be useful for a funny joke or two.
Direct Use
This dataset does not have any captions or phrases supplied with it.
The image column contains a byte string of the JPEG data as read from storage.
Out-of-Scope Use
This data should not be used to generate offensive content.
Dataset Structure
Fields:
- filename (str)
- image hash (str)
- width (int)
- height (int)
- image (bytes, JPEG)
Dataset Creation
Curation Rationale
A LoRA exists that seems to use a similar dataset, but the dataset was not provided or available at any point in time. This dataset is an attempt to reproduce the same results.
Source Data
- User-submitted photographs
Personal and Sensitive Information
Some images may contain faces or identities of individuals. By using this dataset, you agree not to attempt to discover the identity of these people.
Bias, Risks, and Limitations
This dataset's bias is that of its users', as all images were hand-selected for inclusion.