Video-As-Prompt: Unified Semantic Control for Video Generation

  🔥 News

Oct 24, 2025: 📖 We release the first unified semantic video generation model, Video-As-Prompt (VAP)! Oct 24, 2025: 🤗 We release the VAP-Data, the largest semantic-controlled video generation datasets with more than $100K$ samples! Oct 24, 2025: 👋 We present the technical report of Video-As-Prompt, please check out the details and spark some discussion!… See the full description on the dataset page: https://huggingface.co/datasets/BianYx/VAP-Data.

VQA is a multimodal task wherein, given an image and a natural language question related to the image, the objective is to produce a natural language answer correctly as output.

It involves understanding the content of the image and correlating it with the context of the question asked. Because we need to compare the semantics of information present in both of the modalities — the image and natural language question related to it — VQA entails a wide range of sub-problems in both CV and NLP (such as object detection and recognition, scene classification, counting, and so on). Thus, it is considered an AI-complete task.

Context

Action Recognition in video is known to be more challenging than image recognition problems. Unlike image recognition models which use 2D convolutional neural blocks, action classification models require additional dimensionality to capture the spatio-temporal information in video sequences. This intrinsically makes video action recognition models computationally intensive and significantly more data-hungry than image recognition counterparts. Unequivocally, existing video datasets such as Kinetics, AVA, Charades, Something-Something, HMDB51, and UFC101 have had tremendous impact on the recently evolving video recognition technologies. Artificial Intelligence models trained on these datasets have largely benefited applications such as behavior monitoring in elderly people, video summarization, and content-based retrieval. However, this growing concept of action recognition has yet to be explored in Intelligent Transportation System (ITS), particularly in vital applications such as incidents detection. This is partly due to the lack of availability of annotated dataset adequate for training models suitable for such direct ITS use cases. In this paper, the concept of video action recognition is explored to tackle the problem of highway incident detection and classification from live surveillance footage. First, a novel dataset - HWID12 (Highway Incidents Detection) dataset is introduced. The HWAD12 consists of 11 distinct highway incidents categories, and one additional category for negative samples representing normal traffic. The proposed dataset also includes 2780+ video segments of 3 to 8 seconds on average each, and 500k+ temporal frames. Next, the baseline for highway accident detection and classification is established with a state-of-the-art action recognition model trained on the proposed HWID12 dataset. Performance benchmarking for 12-class (normal traffic vs 11 accident categories), and 2-class (incident vs normal traffic) settings is performed. This benchmarking reveals a recognition accuracy of up to 88% and 98% for 12-class and 2-class recognition setting, respectively.

Data Acquisition

The Proposed Highway Incidents Detection Dataset (HWID12) is the first of its kind dataset aimed at fostering experimentation of video action recognition technologies to solve the practical problem of real-time highway incident detections which currently challenges intelligent transportation systems. The lack of such dataset has limited the expansion of the recent breakthroughs in video action classification for practical uses cases in intelligent transportation systems.. The proposed dataset contains more than 2780 video clips of length varying between 3 to 8 seconds. These video clips capture moments leading to, up until right after an incident occurred. The clips were manually segmented from accident compilations videos sourced from YouTube and other videos data platforms.

Content

There is one main zip file available for download. The zip file contains 2780+ video clips. 1) 12 folders
2) each folder represents an incident category. One of the classes represent the negative sample class which simulates normal traffic.

Terms and Conditions

• Videos provided in this dataset are freely available for research and education purposes only. Please be sure to properly credit the authors by citing the article below.
• Be sure to upvote this dataset if you find it useful by scrolling up and clicking the up-Arrow ^ sign at the top banner of the page, next to "New Notebook" button.
• Be sure to blur out all plate numbers before publishing any of the contents available in this dataset.

Acknowledgements

Any publication using this database must reference to the following journal manuscript:

• Landry Kezebou, Victor Oludare, Karen Panetta, James Intriligator, and Sos Agaian "Highway accident detection and classification from live traffic surveillance cameras: a comprehensive dataset and video action recognition benchmarking", Proc. SPIE 12100, Multimodal Image Exploitation and Learning 2022, 121000M (27 May 2022); https://doi.org/10.1117/12.2618943

Note: if the link is broken, please use http instead of https.

In Chrome, use the steps recommended in the following website to view the webpage if it appears to be broken https://www.technipages.com/chrome-enabledisable-not-secure-warning

Other relevant datasets VCoR dataset: https://www.kaggle.com/landrykezebou/vcor-vehicle-color-recognition-dataset VRiV dataset: https://www.kaggle.com/landrykezebou/vriv-vehicle-recognition-in-videos-dataset

For any enquires regarding the HWID12 dataset, contact: landrykezebou@gmail.com

YOLO Version Test Dataset

## Overview

YOLO Version Test Dataset is a dataset for object detection tasks - it contains Objects annotations for 1,992 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

Lefterson/pecp0pv1-dataset dataset hosted on Hugging Face and contributed by the HF Datasets community

The application of Artificial Intelligence (AI) has been evident in the agricultural sector recently. The main goal of AI in agriculture is to improve crop yield, control crop pests/diseases, and reduce cost. The agricultural sector in developing countries faces severe in the form of disease and pest infestation, the knowledge gap between farmers and technology, and a lack of storage facilities, among others. To help address some of these challenges, this work presents crop pests/disease datasets sourced from local farms in Ghana. The dataset is presented in two folds; the raw images which consists of 24,881 images ( 6,549-Cashew, 7,508-Cassava, 5,389-Maize, and 5,435-Tomato) and augmented images which is further split into train and test set consists of 102,976 images (25,811-Cashew, 26,330-Cassava, 23,657-Maize, and 27,178-Tomato), categorized into 22 classes. All images are de-identified, validated by expert plant virologists, and freely available for use by the research community.

Dataset Ow

## Overview

Dataset Ow is a dataset for object detection tasks - it contains Player annotations for 10,000 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

Verified caffeine content for 40+ drinks with serving sizes and per-entry source citations (USDA FoodData Central, published menu data). Updated as sources change.

REU Original Metadataset

## Overview

REU Original Metadataset is a dataset for object detection tasks - it contains Transmission Lines UJFL annotations for 2,485 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

About this dataset

Context

The dataset tabulates the population of Ronceverte by gender, including both male and female populations. This dataset can be utilized to understand the population distribution of Ronceverte across both sexes and to determine which sex constitutes the majority.

Key observations

There is a majority of female population, with 55.02% of total population being female. Source: U.S. Census Bureau American Community Survey (ACS) 2019-2023 5-Year Estimates.

Content

When available, the data consists of estimates from the U.S. Census Bureau American Community Survey (ACS) 2019-2023 5-Year Estimates.

Scope of gender :

Please note that American Community Survey asks a question about the respondents current sex, but not about gender, sexual orientation, or sex at birth. The question is intended to capture data for biological sex, not gender. Respondents are supposed to respond with the answer as either of Male or Female. Our research and this dataset mirrors the data reported as Male and Female for gender distribution analysis. No further analysis is done on the data reported from the Census Bureau.

Variables / Data Columns

• Gender: This column displays the Gender (Male / Female)
• Population: The population of the gender in the Ronceverte is shown in this column.
• % of Total Population: This column displays the percentage distribution of each gender as a proportion of Ronceverte total population. Please note that the sum of all percentages may not equal one due to rounding of values.

Good to know

Margin of Error

Data in the dataset are based on the estimates and are subject to sampling variability and thus a margin of error. Neilsberg Research recommends using caution when presening these estimates in your research.

Custom data

If you do need custom data for any of your research project, report or presentation, you can contact our research staff at research@neilsberg.com for a feasibility of a custom tabulation on a fee-for-service basis.

Inspiration

Neilsberg Research Team curates, analyze and publishes demographics and economic data from a variety of public and proprietary sources, each of which often includes multiple surveys and programs. The large majority of Neilsberg Research aggregated datasets and insights is made available for free download at https://www.neilsberg.com/research/.

Recommended for further research

This dataset is a part of the main dataset for Ronceverte Population by Race & Ethnicity. You can refer the same here

Specific up-to-date data on each of the playable characters in Genshin Impact. Selected data include vision, rarity, release date, character stats and various ascension materials. Other fields may also be added as requested.

Dataset Notes:

All characters belong to HoYoVerse. All character data was scraped from the Genshin Impact Fandom Wiki using the BeautifulSoup package in Python but may still have logical inconsistencies. Note that the .csv file is encoded in the 'UTF-8' format to ensure proper display of special characters.

Dataset updates:

Version 1 (13 Aug 2024): All playable characters available as of game version 4.8, at the end of Fontaine and right before the Natlan release.

VLM-3R Training Data

Training QA data for VLM-3R: vsibench_train/ (VSI-Bench-style tasks) and vstibench_train/ (VSTI-Bench tasks over ScanNet train split).

  Erratum (2026-07-13): corrected camera-position ground truth

A bug in the QA generation pipeline (reported by Jacob Yeung, CMU) extracted the camera center from camera-to-world poses using -R.T @ t instead of pose[:3, 3]. Answers in five vstibench_train files depended on the camera's world position and have… See the full description on the dataset page: https://huggingface.co/datasets/Journey9ni/VLM-3R-DATA.

This dataset contains the digitized treatments in Plazi based on the original journal article Yüceol, Fulya, Çeçen, Ömer, Göktürk, Ramazan Süleyman, Kaya, Ergun, Ağar, Hacer (2024): Valeriana alanyense (Caprifoliaceae): a new species from Southwest Anatolia (Turkey) based on morphological and molecular data. Phytotaxa 653 (1): 79-90, DOI: 10.11646/phytotaxa.653.1.6, URL: http://dx.doi.org/10.11646/phytotaxa.653.1.6

Abstract

Valeriana alanyense a new species to science from Southwestern of Turkey, is described and illustrated. The new species exhibits affinities to V. speluncaria and it differs from it in certain morphological characteristics: leaf shape and size, elongated stems and smaller corolla length. Geographical distribution at lower altitudes also sets it significantly from V. speluncaria. A comprehensive description, photos, information regarding its habitat and ecology, comparisons to related species, a distribution map, and conservation status are additionally considered. Moreover, morphological characterizations of V. alanyense and V. speluncaria were supported by using the ISSR PCR technique in molecular analyses. For molecular analyses, six ISSR primers that gave the most polymorphic band profiles for the Valeriana genus were selected among a total of twenty-five primers. As a result of PCR performed with the six selected ISSR primers, a total of 65 band profiles ranging from 250bp to 1200bp were seored. According to the obtained band profile scores and the constructed dendogram analysis, V. alanyense and V. speluncaria can be separated from each other at the species level. This result supports the data obtained from morphological analysis.

Vector datasets of CWHR range maps are one component of California Wildlife Habitat Relationships (CWHR), a comprehensive information system and predictive model for Californias wildlife. The CWHR System was developed to support habitat conservation and management, land use planning, impact assessment, education, and research involving terrestrial vertebrates in California. CWHR contains information on life history, management status, geographic distribution, and habitat relationships for wildlife species known to occur regularly in California. Range maps represent the maximum, current geographic extent of each species within California. They were originally delineated at a scale of 1:5,000,000 by species-level experts and have gradually been revised at a scale of 1:1,000,000. For more information about CWHR, visit the CWHR webpage (https://www.wildlife.ca.gov/Data/CWHR). The webpage provides links to download CWHR data and user documents such as a look up table of available range maps including species code, species name, and range map revision history; a full set of CWHR GIS data; .pdf files of each range map or species life history accounts; and a User Guide.

This dataset contains the digitized treatments in Plazi based on the original journal article Li, Yu-Long, Jin, Meng-Jie, Zhao, Jian, Liu, Zu-Yao, Wang, Ying-Yong, Pang, Hong (2014): Description of two new species of the genus Megophrys (Amphibia: Anura: Megophryidae) from Heishiding Nature Reserve, Fengkai, Guangdong, China, based on molecular and morphological data. Zootaxa 3795 (4): 449-471, DOI: 10.11646/zootaxa.3795.4.5

Abstract

Two new species, Megophrys acuta sp. nov. and Megophrys obesa sp. nov., are described based on a series of specimens collected from Heishiding Nature Reserve, Fengkai County, Guangdong Province, China. They can be distinguished from other known congeners occurred in southern and eastern China by morphological characters and molecular divergence in the mitochondrial 16S rRNA gene. M. acuta is characterized by small and slender body with adult females measuring 28.1–33.6 mm and adult males measuring 27.1–33.0 mm in snout-vent length; snout pointed, strongly protruding well beyond margin of lower jaw; canthus rostralis well developed and sharp; hindlimbs short, the heels not meeting, tibio-tarsal articulation reaching forward the pupil of eye. M. obesa is characterized by stout and slightly small body with adult females measuring 37.5–41.2 mm, adult male measuring 35.6 mm in snout-vent length; snout round in dorsal view; canthus rostralis developed; hindlimbs short, the heels not meeting, tibio-tarsal articulation reaching forward the posterior margin of eye. The discovery of these two new species further confirms that the diversity of this genus has been significantly underestimated. At present the genus Megophrys contains 56 species of which 35 species are distributed in China.

Key words: Guangdong Province, China, Megophryidae, Megophrys acuta sp. nov., Megophrys obesa sp. nov., mitochondrial DNA, morphology, taxonomy

The Aurora Multi-Sensor Dataset is an open, large-scale multi-sensor dataset with highly accurate localization ground truth, captured between January 2017 and February 2018 in the metropolitan area of Pittsburgh, PA, USA by Aurora (via Uber ATG) in collaboration with the University of Toronto. The de-identified dataset contains rich metadata, such as weather and semantic segmentation, and spans all four seasons, rain, snow, overcast and sunny days, different times of day, and a variety of traffic conditions.
The Aurora Multi-Sensor Dataset contains data from a 64-beam Velodyne HDL-64E LiDAR sensor and seven 1920x1200-pixel resolution cameras including a forward-facing stereo pair and five wide-angle lenses covering a 360-degree view around the vehicle.
This data can be used to develop and evaluate large-scale long-term approaches to autonomous vehicle localization. Its size and diversity make it suitable for a wide range of research areas such as 3D reconstruction, virtual tourism, HD map construction, and map compression, among others.
The data was first presented at the International Conference on Intelligent Robots and Systems (IROS) in 2020, where it was nominated as a Finalist for Best Application Paper at the conference.

The Allen Brain Observatory – Visual Coding is a large-scale, standardized survey of physiological activity across the mouse visual cortex, hippocampus, and thalamus. It includes datasets collected with both two-photon imaging and Neuropixels probes, two complementary techniques for measuring the activity of neurons in vivo. The two-photon imaging dataset features visually evoked calcium responses from GCaMP6-expressing neurons in a range of cortical layers, visual areas, and Cre lines. The Neuropixels dataset features spiking activity from distributed cortical and subcortical brain regions, collected under analogous conditions to the two-photon imaging experiments. We hope that experimentalists and modelers will use these comprehensive, open datasets as a testbed for theories of visual information processing.

This data release contains time-lapse imagery taken at U.S. Geological Survey (USGS) stream gaging stations with associated hydrologic and meteorological data related to each image. These data are to help improve the development of models in detecting water elevation at a given stream gaging station. Images of the water surface and surroundings at USGS stream gaging stations were taken at varying time intervals ranging between every five minutes to an hour. Cameras used include trail cameras, web cameras, and the custom river imagery sensing (RISE) camera. Time-lapse images for each USGS stream gaging station are provided in compressed files (file extension .7z). These files are named in a format to identify the USGS stream gaging station’s site number and station name. Hydrologic and meteorological data including stage, discharge, gage height, water surface elevation (or level), precipitation, relative humidity, water and/or air temperature, dew point, air pressure, and other weather condition information associated with each image are provided in comma-separated values files (file extension .csv). These files are named for users to easily relate the time-lapse images provided in the compressed files to the associated data using the same file naming convention. The hydrologic and meteorological data associated to each image is not the same and varies for each stream gaging station.

CodeXBT/kairos-dataset dataset hosted on Hugging Face and contributed by the HF Datasets community

This dataset contains the digitized treatments in Plazi based on the original journal article Faisal, Mohammad, Singh, Sudhir (2014): Carabid (Coleoptera) type collection at National Forest Insect Collection (NFIC), Forest Research Institute, Dehradun (India). Zootaxa 3786 (3): 331-358, DOI: 10.11646/zootaxa.3786.3.5

Abstract

Members of family Carabidae (Insecta: Coleoptera) are a dominant group of terrestrial predators. National Forest Insect Collection (NFIC) of Forest Research Institute, Dehradun (India) has a good collection of carabids rich in type material. Here we report the details of the type specimens of 139 species included in 49 genera, 24 tribes and 14 subfamilies. Colour automontaged photographs of each type along with its original labels are also included.

Key words: Holotype, co-types, updation, digitization, Auto-montage 3 - D imaging system