datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
fight-detection-video
Fight Dataset - 1,000+ videos
The dataset comprises 1,000 high-quality videos captured from static and moving surveillance cameras at resolutions up to 1920×1080 and 30 FPS. These video clips contain fights and violent actions recorded in controlled environments. It is designed for research in violence detection, focusing on various detection techniques and algorithms, primarily aimed at advancing public safety and security systems.
By utilizing this dataset, researchers and… See the full description on the dataset page: https://huggingface.co/datasets/UniDataPro/fight-detection-video.hispanic-people-liveness-detection-video-dataset
Biometric Attack Dataset, Hispanic People
The similar dataset that includes all ethnicities - Anti Spoofing Real Dataset
The dataset for face anti spoofing and face recognition includes images and videos of hispanic people. 32,600+ photos & video of 16,300 people from 20 countries. The dataset helps in enchancing the performance of the model by providing wider range of data for a specific ethnic group.
The videos were gathered by capturing faces of genuine individuals… See the full description on the dataset page: https://huggingface.co/datasets/UniqueData/hispanic-people-liveness-detection-video-dataset.asian-people-liveness-detection-video-dataset
Biometric Attack Dataset, Asian People
The similar dataset that includes all ethnicities - Anti Spoofing Real Dataset
The dataset for face anti spoofing and face recognition includes images and videos of asian people. 30,600+ photos & video of 15,300 people from 32 countries. All people presented in the dataset are South Asian, East Asian or Middle Asian. The dataset helps in enchancing the performance of the model by providing wider range of data for a specific ethnic… See the full description on the dataset page: https://huggingface.co/datasets/UniqueData/asian-people-liveness-detection-video-dataset.black-people-liveness-detection-video-dataset
Biometric Attack Dataset, Black People
The similar dataset that includes all ethnicities - Anti Spoofing Real Dataset
The dataset for face anti spoofing and face recognition includes images and videos of black people. The dataset helps in enchancing the performance of the model by providing wider range of data for a specific ethnic group.
The videos were gathered by capturing faces of genuine individuals presenting spoofs, using facial presentations. Our dataset proposes… See the full description on the dataset page: https://huggingface.co/datasets/UniqueData/black-people-liveness-detection-video-dataset.rav4-video-detections
RAV4 Video Detection and Retrieval Index
This dataset contains two Parquet files produced by a YOLO26 object detection pipeline
applied to a Toyota RAV4 exterior review video. Together they form a complete
image-to-video semantic retrieval system.
Source Video
Field
Value
Video ID
YcvECxtXoxQ
YouTube URL
https://www.youtube.com/watch?v=YcvECxtXoxQ
Sampling rate
0.5 fps (one frame every 2 seconds)
Total frames
~1,396
Detector… See the full description on the dataset page: https://huggingface.co/datasets/ccczy-czy/rav4-video-detections.assignment2-video-detections
Assignment 2 – Image-to-Video Semantic Retrieval
This dataset contains a detection index (Parquet) for a segment of a YouTube video.
Video
video_id: YcvECxtXoxQ
segment processed: 18:39 to 25:04
sampling: 1 frame/second
timestamp offset applied: 1119 seconds (timestamps match the original full video)
Detector
YOLOv8 segmentation (Ultralytics)
base: yolov8n-seg.pt
fine-tuned on: carparts-seg
confidence threshold: 0.25
Files… See the full description on the dataset page: https://huggingface.co/datasets/divyanshmaurya1/assignment2-video-detections.fight-detection-video
Fight Dataset - 1,000+ videos
This dataset contains 1,000 high-quality videos of simulated physical altercations recorded in controlled environments, captured from static and moving surveillance camera views at up to 1920×1080 resolution and 30 FPS. Designed for violence detection, action recognition, and public safety systems, this surveillance dataset includes rich metadata annotations enabling accurate camera fight analysis and training violence detection models.- Get the data… See the full description on the dataset page: https://huggingface.co/datasets/ud-smart-city/fight-detection-video.Auditorium-Graduation-Ceremony-Human-Pose-Detection-Video
Auditorium Graduation Ceremony Human Pose Detection Video
In the current fields of video surveillance and behavior recognition, accurately detecting and analyzing human poses is a key challenge. Existing pose detection technologies perform poorly in complex backgrounds and multi-person environments, and require high real-time performance. This dataset aims to address the need for accurately detecting and recognizing various human poses and behaviors in auditoriums, schools, and… See the full description on the dataset page: https://huggingface.co/datasets/Mobiusi/Auditorium-Graduation-Ceremony-Human-Pose-Detection-Video.opensource-video-vehicle-detectionvideo_detections
Dataset Card for video_detections
Parquet dataset containing detections from YOLO model
Dataset Details
Column Name
Data Type
Description
video_id
string
the id of the source video
frame_index
int64
the frame number where the detection occured
timestamp
float64
the time (in seconds) in the video where the detection occured
class_label
string
the object class of the detection
bounding_box
list[float]the bounding box coordinates [x1, y1, x2, y2].… See the full description on the dataset page: https://huggingface.co/datasets/Anand-Rao745/video_detections.rav4-video-detections
RAV4 Exterior Video Detections
This dataset contains object-part detections for the YouTube video corpus YcvECxtXoxQ (car exterior).
Files
video_detections.parquet: one row per detection.
Parquet schema
video_id (string): YouTube video id (YcvECxtXoxQ)
frame_index (int): extracted frame number (from frame_XXXXXX.jpg)
timestamp_sec (int): time in seconds in the source video
class_label (string): predicted exterior part class (e.g., wheel, front_bumper)… See the full description on the dataset page: https://huggingface.co/datasets/PotriAbhiB/rav4-video-detections.rav4-video-detections
RAV4 Video Detections
Car-part detections and retrieval results from the Toyota RAV4 exterior review video (YcvECxtXoxQ) using a two-pass smart sampling strategy:
Coarse pass (1 frame / 5 s) scored by CLIP zero-shot exterior classifier (openai/clip-vit-base-patch32).
Dense pass (1 fps) applied only to windows with exterior confidence >= 0.70.
YOLOv8n-seg fine-tuned on the Ultralytics Car Parts Segmentation dataset runs on all dense frames.
File 1:… See the full description on the dataset page: https://huggingface.co/datasets/Bineet2407/rav4-video-detections.midjourney-video-detection-complete
🎬 Midjourney Video Detection Dataset (Complete)
📝 Description
Ce dataset contient 49 vidéos générées par Midjourney avec accès direct aux fichiers vidéo pour l'entraînement de modèles de détection de contenu vidéo généré par IA.
✅ Caractéristiques
🎥 49 vidéos MP4 entièrement uploadées (221.6 MB)
📊 Métadonnées enrichies avec descriptions détaillées
🔍 Accès direct aux vidéos via le format VideoFolder
📱 Compatible Dataset Viewer Hugging Face
🚀 Prêt pour… See the full description on the dataset page: https://huggingface.co/datasets/zelus82/midjourney-video-detection-complete.drone_video_detections
Output Tracking Videos
drone_video_1_tracked.mp4
drone_video_2_tracked.mp4
Dataset Choice and Detector Configuration
For this project I used the Drone detection Computer Vision Model dataset from Roboflow to train the YOLO model. I chose this dataset because it contains 991 images of drones taken from many different angles and in many different environments which improves generalization.
I used the YOLO model for object detection. I chose YOLO… See the full description on the dataset page: https://huggingface.co/datasets/Anand-Rao745/drone_video_detections.video-detections
Video Semantic Index Dataset
Overview
This dataset contains frame-level object detection results extracted from sampled video frames.
Detections were generated using a YOLO-based object detection model and stored in Apache Parquet format for efficient structured indexing and downstream semantic retrieval.
Each row represents a single detected object instance.
File Information
Filename: video_semantic_index.parquet
Format: Apache Parquet
Compression: SNAPPY… See the full description on the dataset page: https://huggingface.co/datasets/ntdky/video-detections.drone-video-detections
Drone Video Detections
Overview
This dataset contains object detections from two drone surveillance videos,
serving as the detection index for a Kalman filter tracking system.
Detector: YOLOv8n fine-tuned on DRONES_NEW-4 dataset (mAP50=0.970 || mAP50-95=0.71)
Total detections: 918Confidence threshold: 0.39
Video
Frames
Frames with Detections
video1
828
507
video2
2,580
393
Schema
Column
Type
Description
video_id
string
Video… See the full description on the dataset page: https://huggingface.co/datasets/AhmedJaheen/drone-video-detections.concrete_circles_change_detection_video
About Dataset
The dataset consists of 12 high-resolution videos of 3840 × 2160. The videos contain rotating concrete cylinders with a diameter of 22mm. Each video contains two parts separated by a few seconds of black screen, where each part corresponds to the same concrete sample being filmed under different lighting conditions. The video recording process is described below:
Concrete sample is fixed to the installation so that the upper base is at the height indicated by the… See the full description on the dataset page: https://huggingface.co/datasets/georgiisirotenko/concrete_circles_change_detection_video.assignment2-video-detections
Assignment 2 – Image-to-Video Semantic Retrieval
Overview
This dataset contains object detection results extracted from the YouTube video:
https://www.youtube.com/watch?v=YcvECxtXoxQ
The goal of this assignment was to build a semantic retrieval system that can identify video segments where a specific car exterior component appears.
Instead of manually labeling timestamps, the system detects objects in sampled video frames and stores structured detection results. These… See the full description on the dataset page: https://huggingface.co/datasets/TejaAlapati/assignment2-video-detections.h2-video-detections
Video Detections and Query Clips
Parquet outputs for the video detection and image semantic search pipeline (car-parts detector on video + RAV4 query images).
Files
video_detections.parquet — One row per video frame; each row has a list of object detections for that frame.
query_longest_clips.parquet — One row per query image; each row has the longest contiguous video clip where the detected car parts appear, with a YouTube embed URL.
Schema… See the full description on the dataset page: https://huggingface.co/datasets/rushilara/h2-video-detections.rav4-video-detections
Assignment 2: Image-to-Video Semantic Retrieval
Dataset URL: https://huggingface.co/datasets/saks2002/rav4-video-detections
Assignment 2 Report: Image-to-Video Semantic Retrieval
1. Detector Choice and Configuration
For this assignment, I chose the YOLOWorld-l (Large) model for object detection.
Why YOLOWorld?
The assignment required detection at the "object part level" (e.g., hoods, wheels, headlights). While standard YOLOv8 models (trained on… See the full description on the dataset page: https://huggingface.co/datasets/saks2002/rav4-video-detections.car-parts-video-detections
Car Parts Video Detections
Detection index for a car exterior review video. One row per detected car part per frame.
Video: YouTube — YcvECxtXoxQDetector: YOLOv8n-seg fine-tuned on carparts-segSampling: 1 fps
Schema
Column
Type
Description
video_id
string
YouTube video ID
frame_index
int
Frame number (1-based)
timestamp
float
Seconds into video
class_label
string
Part name (e.g. front_bumper)
class_id
int
Class number
bbox_x_min, bbox_y_min… See the full description on the dataset page: https://huggingface.co/datasets/Masutarisu/car-parts-video-detections.sonar_detection_videorav4-video-detections
RAV4 Exterior Video Detections
Overview
This dataset contains object detections from a Toyota RAV4 2026 exterior review video.
It serves as the retrieval index for an image-to-video semantic search system.
Source video: YouTube - YcvECxtXoxQDetector: YOLOv8n-seg fine-tuned on the Ultralytics carparts-seg dataset (mAP50=0.666)Frames sampled: 1397 frames at 1 frame per 2 secondsTotal detections: 3,489
Schema
Column
Type
Description
video_id… See the full description on the dataset page: https://huggingface.co/datasets/AhmedJaheen/rav4-video-detections.Video_DetectionsEach row in the dataset corresponds to one detected object instance in a single frame, along with spatial and temporal metadata.
Detection & Temporal Indexing
The video was sampled at a constant rate (5 seconds per frame).
Each sampled frame is assigned:
A frame index (sequential order)
A timestamp in seconds indicating its position in the video.
YOLO was applied to each frame to produce:
Bounding box coordinates
Component class labels
Confidence scores
This temporal… See the full description on the dataset page: https://huggingface.co/datasets/Harindham/Video_Detections.rav4-video-detection
RAV4 Video Detections
Object detection index and image-to-video retrieval results for the Toyota RAV4 exterior video (YcvECxtXoxQ).
Files
File
Description
video_detections.parquet
All car-part detections from the video corpus (92 rows)
retrieval_results.parquet
Retrieval results matching query images to video segments (72 rows)
Schema: video_detections.parquet
Column
Type
Description
video_id
string
YouTube video ID… See the full description on the dataset page: https://huggingface.co/datasets/nazzzz5265/rav4-video-detection.rav4-video-detections
RAV4 Video Detections
This dataset contains object detection results from a Toyota RAV4 exterior review video (YouTube ID: YcvECxtXoxQ), used for image-to-video semantic retrieval.
Files
video_detections.parquet — Frame-level detections from the input video
retrieval_results.parquet — Query-to-video retrieval results
Schema: video_detections.parquet
Column
Type
Description
video_id
string
YouTube video ID
frame_index
int
Frame number… See the full description on the dataset page: https://huggingface.co/datasets/ZuhairMunawar/rav4-video-detections.assignment2-video-detections
Assignment 2 – Video Detections Index
This repository contains the offline detection index and retrieval results
for the video-to-query retrieval task.
1. Video Detections (Required)
File: video_detections.parquet
Each row corresponds to one object detection on one sampled frame.
Schema
video_id (string): YouTube video ID
frame_index (int): 0-based index of sampled frame
timestamp_sec (int): frame timestamp in seconds
class_label (string): predicted car… See the full description on the dataset page: https://huggingface.co/datasets/Congque/assignment2-video-detections.rav4-video-detections
RAV4 Video Object Detections
This repository contains the detection output parquet file for Assignment 2:
"Image-to-Video Semantic Retrieval via Object Detection".
File
yolo_detections_clip.parquet
Each row represents one detection from a sampled video frame.
Notes
The parquet file is used as the interface between detection and retrieval.
Main fields
video_id
detector_name
frame_index / timestamp
class_label
bounding_box
confidence_score
video-detectionsclass_label
bounding_box
confidence_score
CS370_Assignment_2_Video_Detections
Description
This dataset corresponds to car part detections made from a Toyota RAV4 review video (https://www.youtube.com/watch?v=YcvECxtXoxQ).
Data Schema
video_id: Name of the video where detection occurred
frame_index: Relative index of frame where detection occurred
timestamp: Time (MM:SS) in the video at which detection occurred
class_id: ID of detected car part
class_label: Name of detected car part
bounding_box: Corners (X-Coordinate Bottom Left, Y-Coordinate… See the full description on the dataset page: https://huggingface.co/datasets/aakashk99/CS370_Assignment_2_Video_Detections.