Monocular Household Egocentric Video Dataset
Egocentric & First-Person Data
Tags and Keywords
$7,800
About
About
A collection of 610 hours of real-world monocular egocentric (first-person) video recordings capturing human activities in household and residential environments.
The dataset provides demonstrations of everyday human actions, object interactions and domestic workflows from a first-person perspective.
Designed for Home Robotics, Physical AI, Embodied AI, Computer Vision, Vision-Language Models (VLMs) and imitation learning, the footage supports the development of AI systems capable of understanding human behavior and performing everyday tasks.
Available as an off-the-shelf (OTS) dataset under non-exclusive commercial licensing, with flexible volume selection and custom delivery.
Data Product Features
1. Monocular First-Person Recordings
- Real-world, human-performed household activities
- First-person RGB video recordings
- Single-camera (monocular) perspective
- Typically Full HD (1920 × 1080) at 30 FPS, subject to recording specifications
- Natural human-object and human-environment interactions
- Authentic residential settings rather than simulated environments
2. Household Activity Coverage
The dataset captures everyday domestic activities and human interactions with household objects, tools and workspaces.
Specific task categories, activity descriptions and recording volumes can be provided from the available inventory upon request.
3. Human Demonstrations
Recordings provide visual information relevant to:
- Human hand and object interactions
- Task execution and activity sequences
- Everyday object handling
- Human movements in domestic environments
- Understanding household workflows
4. Metadata & Annotations
Depending on the selected recordings, available information may include:
- Activity and task descriptions
- Task categories
- Recording duration
- Video specifications
- Session-level information
- Action labels and annotations
- IMU motion data, where available
Metadata, annotation coverage and sensor availability vary between recordings and are confirmed before delivery.
Distribution
- Data Volume: 610 hours
- Video Type: Monocular RGB, first-person
- File Format: Primarily MP4, subject to source files
- Resolution: Typically 1920 × 1080 (Full HD)
- Frame Rate: Typically 30 FPS
- Recording Perspective: Human-worn, first-person
- Recording Environment: Household and residential settings
- Dataset Organization: By task category, activity and recording session, where available
- Metadata: Task and recording information for selected footage
- Annotations: Available for selected recordings
- Delivery Method: Custom secure cloud delivery
- Dataset Selection: Flexible by available activity category and required volume
Exact file counts, recording durations, storage sizes and metadata coverage will be confirmed before delivery.
Usage
This dataset is suitable for:
- Home Robotics: Training models to understand human activities and everyday household interactions.
- Imitation Learning: Learning action sequences from real-world human demonstrations.
- Physical AI: Training AI models to interpret physical tasks, objects and environments.
- Embodied AI: Developing systems that understand human behavior and interactions with household objects.
- Vision-Language Models: Video understanding, activity descriptions and multimodal AI training.
- Computer Vision: Human action recognition, object interaction understanding and activity classification.
- Robotic Manipulation Research: Studying demonstrations of everyday object handling and task execution.
- AI Model Evaluation: Evaluating model understanding of real-world household activities.
Coverage
Geographic Coverage: Asia, subject to verification of recording locations.
Environment Coverage: Real-world household and residential environments.
Activity Coverage: Everyday household tasks, domestic activities and human-object interactions, as documented in the available recordings.
Collection Period: Varies by recording session.
Participants: Human operators performing everyday household activities.
Data Type: Real-world human demonstrations rather than synthetic or simulated footage.
Excluded Categories: Industrial, commercial, agricultural and transportation activities are offered separately.
License
Proprietary – Non-Exclusive Commercial Licensing
The dataset is offered under non-exclusive commercial licensing, allowing TRANSFORM Solutions to license the same recordings to multiple customers.
Permitted AI training, fine-tuning, evaluation and derivative uses are governed by the applicable licensing agreement and verified underlying rights.
Resale, redistribution or unauthorized sharing of the original video recordings is prohibited.
Dataset volume, permitted usage, pricing and delivery terms are agreed before purchase.
Who Can Use It
- Home Robotics Companies: Developing robots capable of understanding household activities.
- Physical AI Developers: Training models on real-world human actions and domestic environments.
- Embodied AI Research Teams: Human demonstration learning and physical-world understanding.
- Computer Vision Companies: Activity recognition and human-object interaction modeling.
- Vision-Language Model Developers: Household video understanding and multimodal model training.
- AI Research Laboratories: Research into imitation learning, robotics and human behavior.
- Universities: Academic research in robotics, computer vision and AI.
Data Dictionary
The following fields describe dataset attributes and metadata potentially available for selected recordings. Not all fields are guaranteed for every video.
| Column Name | Data Type | Description |
|---|---|---|
| dataset_id | String | Dataset or collection identifier |
| dataset_category | String | Household activity category |
| task_category | String | Task grouping or activity family |
| task_description | Text | Description of recorded household activity |
| tools_objects | Text | Objects or tools involved, where documented |
| video_duration | Numeric | Recording duration in seconds |
| video_resolution | String | Video resolution |
| frame_rate | Numeric | Frames per second |
| camera_type | String | Monocular RGB / first-person |
| capture_device | String | Recording hardware, where documented |
| recording_environment | String | Household or residential setting |
| session_id | String | Recording session identifier, where available |
| imu_available | Boolean | Availability of IMU data |
| action_labels | String/List | Activity annotations, where available |
| recording_location | String | Geographic information, where documented |
| license_type | String | Non-exclusive commercial |
Additional Information
TRANSFORM Solutions offers off-the-shelf monocular household egocentric datasets for Home Robotics, Physical AI, and commercial AI model development.
Key Highlights:
- 610 hours of household egocentric video
- Real-world human demonstrations
- Monocular first-person RGB recordings
- Authentic domestic environments
- Everyday human-object interactions
- Task metadata and annotations for selected recordings
- Non-exclusive commercial licensing
- Flexible dataset selection and volume-based pricing
- Representative samples available upon request
- Custom secure delivery
Stereo egocentric datasets and industrial/commercial activity recordings are offered separately.
Contact TRANSFORM Solutions for detailed inventory information, samples, technical specifications, and commercial quotations.
TRUSTLoading...
$7,800
Download Dataset in VIDEO Format
Recommended Datasets
Loading recommendations...