Sign in to view source links and access this dataset
Description
ImageNet-R(endition) contains images of art, cartoons, deviantart, graffiti, embroidery, graphics, origami, paintings, patterns, plastic objects, plush objects, sculptures, sketches, tattoos, toys, and video game renditions. The dataset is hosted on Hugging Face by user 'axiong' and was last updated on June 19, 2024. It is constructed from a source file provided by an official implementation to facilitate the evaluation of various pretraining models.
Use Cases
Evaluating model robustness to domain shifts based on the described artistic renditions.
Benchmarking pretrained image classifiers on non-photorealistic data based on categories like cartoons and paintings.
Training or fine-tuning models for style-invariant recognition based on the diverse rendition types mentioned.
Strengths
Dataset is specifically constructed for evaluating pretraining models, indicating a clear purpose.
Contains a diverse range of rendition types as described, including art, cartoons, graffiti, and sculptures.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
Source
Constructed from a source file provided by an official implementation, likely related to the original ImageNet-R research.
Collection Method
The description suggests it was assembled for evaluation purposes, but the specific gathering method is not detailed.
Freshness
Last updated 2024-06-19 06:50:57; freshness should be verified.