Egocentric Video Data Collection
Budget: ₹400 – ₹750 INR
Project Overview:
We are collecting egocentric (first-person POV) video data of everyday household tasks recorded in real residential environments. These videos will be used for human-to-robot / humanoid training datasets.
Recording Requirements (Mandatory):
First-person (POV) recording using a head-mounted smartphone or equivalent device
Recording must reflect natural human viewpoint
Handheld, tripod, or surface-mounted recordings are NOT allowed
Video Specifications:
Resolution: Minimum 1920×1080 (1080p)
Orientation: Landscape only
Frame Rate: 30 FPS minimum (60 FPS preferred)
No slow motion
No vertical videos
Visibility & Framing Rules:
Hands and objects must be clearly visible at all times
Camera angle slightly downward
No blurred or out-of-focus footage
No faces visible (including mirrors, photos, reflections)
No IDs, documents, personal data
Content Rules:
Minimum video duration: 4–5 minutes per video
No idle time longer than 3 seconds
Only meaningful household tasks (kitchen, living room, bedroom, bathroom.)
No eating, drinking, washing hands, sitting idle, or unrelated activities
Editing Rules:
Raw video files only accepted
No trimming
No filters
No overlays, effects, or text, audio
Rejection Criteria - Videos will be rejected if they include:
Vertical videos
Slow-motion recording
Phone held by hand or placed on a surface
Camera zoomed in too much
Blurry, unfocused, or unstable footage
Idle time longer than 3 seconds
Edited or trimmed videos
Eligibility:
Open to international freelancers
Must have access to a residential home environment
Ability to follow strict technical guidelines is required
Before Applying - Applicants must confirm:
They have a head-mounted recording setup
Their device supports 1080p @ 30 FPS and above.
They agree to follow all recording rules strictly
Deliverables:
Raw video files only
Uploaded to assigned cloud folder, sorted by environment
Payment Details:
Payment only for approved hours and rejected videos hours are not paid.
We are collecting egocentric (first-person POV) video data of everyday household tasks recorded in real residential environments. These videos will be used for human-to-robot / humanoid training datasets.
Recording Requirements (Mandatory):
First-person (POV) recording using a head-mounted smartphone or equivalent device
Recording must reflect natural human viewpoint
Handheld, tripod, or surface-mounted recordings are NOT allowed
Video Specifications:
Resolution: Minimum 1920×1080 (1080p)
Orientation: Landscape only
Frame Rate: 30 FPS minimum (60 FPS preferred)
No slow motion
No vertical videos
Visibility & Framing Rules:
Hands and objects must be clearly visible at all times
Camera angle slightly downward
No blurred or out-of-focus footage
No faces visible (including mirrors, photos, reflections)
No IDs, documents, personal data
Content Rules:
Minimum video duration: 4–5 minutes per video
No idle time longer than 3 seconds
Only meaningful household tasks (kitchen, living room, bedroom, bathroom.)
No eating, drinking, washing hands, sitting idle, or unrelated activities
Editing Rules:
Raw video files only accepted
No trimming
No filters
No overlays, effects, or text, audio
Rejection Criteria - Videos will be rejected if they include:
Vertical videos
Slow-motion recording
Phone held by hand or placed on a surface
Camera zoomed in too much
Blurry, unfocused, or unstable footage
Idle time longer than 3 seconds
Edited or trimmed videos
Eligibility:
Open to international freelancers
Must have access to a residential home environment
Ability to follow strict technical guidelines is required
Before Applying - Applicants must confirm:
They have a head-mounted recording setup
Their device supports 1080p @ 30 FPS and above.
They agree to follow all recording rules strictly
Deliverables:
Raw video files only
Uploaded to assigned cloud folder, sorted by environment
Payment Details:
Payment only for approved hours and rejected videos hours are not paid.
Related categories:
Video Services
Videography
Video Production
Filmmaking
Video Processing
Data Collection