wisdomik
wisdomik/QuiltNet-B-32
- zero-shot-image-classification - clip - vision - language - histopathology - histology - medical librarytag: openclip - src: >- https://qu...
Model Documentation
QuiltNet-B-32 Description
QuiltNet-B-32 is a CLIP ViT-B/32 vision-language foundation model trained on the Quilt-1M dataset curated from representative histopathology videos. It can perform various vision-language processing (VLP) tasks such as cross-modal retrieval, image classification, and visual question answering. QuiltNet establishes new state of the art in a wide range of standard datasets, and substantially outperforms prior VLP approaches:
Citation
bibtex
@misc{ikezogwo2023quilt1m,
title={Quilt-1M: One Million Image-Text Pairs for Histopathology},
author={Wisdom Oluchi Ikezogwo and Mehmet Saygin Seyfioglu and Fatemeh Ghezloo and Dylan Stefan Chan Geva and Fatwir Sheikh Mohammed and Pavan Kumar Anand and Ranjay Krishna and Linda Shapiro},
year={2023},
eprint={2306.11207},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
Uses
As per the original OpenAI CLIP model card, this model is intended as a research output for research communities. We hope that this model will enable researchers to better understand and explore zero-shot, arbitrary image classification. We also hope it can be used for interdisciplinary studies of the potential impact of such model.
The OpenAI CLIP paper includes a discussion of potential downstream impacts to provide an example for this sort of analysis.
Direct Use
Zero-shot image classification, image and text retrieval, among others.
Downstream Use
Image classification and other image task fine-tuning, linear probe image classification, image generation guiding and conditioning, among others.
Intended Use
The model is intended as a research output for research communities. We hope that this model will enable researchers to better understand and explore zero-shot, arbitrary image classification. We also hope it can be used for interdisciplinary studies of the potential impact of such models.
#
Primary intended uses
The primary intended users of these models are AI researchers.
We primarily imagine the model will be used by researchers to better understand robustness, generalization, and other capabilities, biases, and constraints of computer vision histopathology models.
Out-of-Scope Use Cases
Any deployed use case of the model
Since the model has not been purposefully trained in or evaluated on any languages other than English, its use should be limited to English language use cases.
Further the above notice, the Quilt-1M dataset used in training of these models has additional considerations, see below.
Training Data
This model was trained with QUILT-1M is an image-text dataset for histopathology. Curated from educational videos on Youtube QUILT-1M contributes the largest dataset for vision language modeling in histopathology.
IMPORTANT NOTE: The motivation behind dataset creation is to democratize research and experimentation around large-scale multi-modal model training and handling of uncurated, large-scale histopathology datasets crawled from publically available internet. Our recommendation is therefore to use the dataset for research purposes.
Evaluation
Evaluation done with code in the CLIP Benchmark suite and results can be found in the paper on a list of varying histology tasks and datasets.