datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
detailedBG_LoraDetailVerifyBench
DetailVerifyBench
Project Page | Paper | GitHub
DetailVerifyBench is a rigorous benchmark designed for dense hallucination localization in long image captions. It comprises 1,000 high-quality images across five distinct domains: Chart, Movie, Nature, Poster, and UI. With an average caption length of over 200 words and dense, token-level annotations of multiple hallucination types, it stands as a challenging benchmark for evaluating the precise hallucination localization capabilities… See the full description on the dataset page: https://huggingface.co/datasets/zyxhhnkh/DetailVerifyBench.DetailVariationsV1
Image Detail Manipulation Dataset
Dataset Summary
Contains sets of images designed for tasks involving controlled manipulation of image details or styles. Each set consists of one input image (representing a baseline detail level, 'f5') and nine corresponding edited versions ('f0' through 'f9'), each representing a different level detail.
The Dataset was realized using SDXL Upscaling and Refiner.
Around 60% of the images used to create this Dataset come from third party… See the full description on the dataset page: https://huggingface.co/datasets/Scaryplasmon96/DetailVariationsV1.flickr8k-turkish-detailed-captionsDetailed captions were genereted by gpt-4o-mini using OpenAI API.
M. E. Unal, B. Citamak, S. Yagcioglu, A. Erdem, E. Erdem, N. Ikizler Cinbis and R. Cakici. TasvirEt: Görüntülerden Otomatik Türkçe Açıklama Oluşturma İçin Bir Denektaşı Veri Kümesi (TasvirEt: A Benchmark Dataset for Automatic Turkish Description Generation from Images). 24. IEEE Sinyal İşleme ve İletişim Uygulamaları Kurultayı (SIU 2016), Zonguldak, Mayis 2016
cache_detailed_C034cache_detailed_C379midjourney-detailed-prompts
Midjourney: Detailed Prompts
This dataset is my attempt in providing a high quality text-to-image dataset with detailed and several levels of prompting for images.
Hope it helps anyone in his research ^^
Thanks goes to ...
midjourney-images dataset
Qwen-VL-Max for descriping images in huge detail.
Command R for long & short prompt generation
povarenok_recipes_detail
povarenok_recipes_detail
Crawled detailed recipes from povarenok.ru website.
Structure
WIP
cache_detailed_C159cache_detailed_C344cache_detailed_C396cache_detailed_C022cache_detailed_C185cache_detailed_C345cache_detailed_C405cache_detailed_C132cache_detailed_C294Image-Detailed-Description-Korean
Image-Detailed-Description-Korean
LLaVA-NeXT에 적혀있는 내용중 High-Quality Knowledge Learning부분에 다음의 내용이 있습니다:
Enhanced Performance with Recaptioned Data
Models trained with recaptioned data (ReCap) datasets, show a trend of enhanced performance in tasks requiring detailed image descriptions and document understanding.
The regenerated captions, ranging from 118K to 3M, demonstrate better scaling behaviors than the original captions, consistently improve model performance across… See the full description on the dataset page: https://huggingface.co/datasets/Nagase-Kotono/Image-Detailed-Description-Korean.Detailed_CaptionGithub|Paper
VerboVision-Detailv2.0cache_detailed_C168cache_detailed_C183cache_detailed_C057cache_detailed_C248cache_detailed_C351detail_cadmodwell-bedbathlrkitch-details-01cache_detailed_C068cache_detailed_C352cache_detailed_C089