Loading...
Loading...
Image-text pairs, instruction tuning, visual QA, cross-modal data, foundation model training data
1,948 datasets
A dataset titled 'test-images-vqa' is hosted on Kaggle. The dataset likely contains images paired with questions and answers for visual question answering tasks. Metadata such as size, columns, and license are currently unknown.
BLIP_test is a dataset hosted on Kaggle. Its title suggests it is likely related to the BLIP (Bootstrapping Language-Image Pre-training) model, a vision-language framework. The dataset's specific content, size, and structure are unknown from the provided metadata.
llava-annotations-pascal-voc is a dataset hosted on Kaggle. The title suggests it contains annotations generated by the LLaVA (Large Language and Vision Assistant) model for the PASCAL VOC object detection and segmentation benchmark. The dataset likely provides question-answer pairs or descriptive labels for images, linking visual content with language.
A collection of training and evaluation files derived from the MultiVENT 2.0 benchmark for text-to-video retrieval. The dataset provides structured query-video pairs within training_data.json designed to facilitate explicit reasoning over video content for relevance assessment.
PhoStream contains 5,572 open-ended QA pairs derived from 578 videos across 4 scenarios and 10 capabilities, released by lucky-lance in 2026. This benchmark evaluates omnimodal assistants in mobile-centric streaming environments, focusing on both on-screen and off-screen phone usage. It specifically tests a model's ability to determine both the timing and the content of responses while processing continuous audio-visual streams.
Highlighting pre-rendered 3D multi-room environments categorized for evaluating spatial reasoning in Vision Language Models. It provides structured visual scene data to support the Theory of Space benchmark, focusing on active exploration and the construction of spatial beliefs.
Kaggle hosts the MiniVLM dataset, which is likely related to vision-language modeling. The dataset's specific content, size, and creation details are not provided in the available metadata.
A dataset titled 'bmps_vlm' published on Kaggle. The title suggests it is likely related to vision-language models, a subfield of multimodal AI. No further metadata is available to confirm its specific contents, size, or origin.
EngVQA-GRPO is a dataset hosted on Kaggle. The title suggests it likely contains English-language visual question answering data. The dataset's specific content, size, and origin are unknown from the provided metadata.
gvlma is an R package for the global validation of linear model assumptions, authored by Edsel A. Pena and Elizabeth H. Slate. The dataset likely contains statistical test results and diagnostic metrics for assessing model fit. It is sourced from the paperswithcode platform, which aggregates resources for the computer science and mathematics communities.
236 real classroom sessions provide data across 5 modalities and 25 features. The dataset is designed for human-AI collaborative analysis of teaching quality. Its origin and creation date are unknown.
BEAT2 Aligned Multimodal Features is a dataset hosted on Kaggle. The title suggests it contains features extracted from multiple data modalities that have been aligned. The dataset's specific content, size, and provenance are unknown.
IMDB multimodal likely contains data from the Internet Movie Database, combining multiple types of media. The dataset is published on Kaggle, but its specific content, size, and creation details are unknown. Its last update date and authorship are not provided.
ROCO Multimodal 4 Clusters Dataset is a dataset hosted on Kaggle. The title suggests it contains multimodal data organized into four clusters. The dataset likely contains data from multiple modalities, such as images and text, intended for clustering tasks.
A multimodal dataset from Kaggle, likely containing data organized into four clusters. The dataset's title suggests it may combine different data types such as images and text. Specific details regarding its size, creation date, and authorship are not provided in the available metadata.
A script for streaming large language model training, authored by uv-scripts and last updated in January 2026. It demonstrates training a Qwen model on Latin using 1.47 million texts streamed directly from the FineWeb-2 dataset on Hugging Face Hub. The associated blog post details the method for training on massive datasets without local downloads.
A Chinese-language dataset of 595,000 items, created by machine-translating the LLaVA-CC3M-Pretrain-595K dataset. The data was uploaded by author 'cyberlangke' to Hugging Face and last updated on 2026-02-25. The description notes the translations are unverified and may contain errors.
107,821 wine bottle images scraped from wine retailer websites, linked to a companion text dataset. The dataset was created by cipher982 and last updated on January 21, 2026. It is intended for computer vision and multimodal machine learning tasks.
Multimodal images for AI-based fall recognition, published on Kaggle. The dataset likely contains visual data intended for training models to detect falls in elderly individuals. Specific details on volume, collection method, and authorship are not provided in the available metadata.
400 million diverse image-caption pairs collected from the open web support multimodal AI research. The dataset includes rich metadata for filtering and deduplication. It was created by Ajax102 and last updated on Hugging Face in January 2026.