Image Data Collection Project - SoftAge AI
Last Updated: 11th March 2026
Download the App
Access the image collection platform using the Play Store link below. You must use this app to
capture all images for this project.
Android:
[Link]
are
iOS:
[Link]
Contributor Guidelines & Overview
This project involves capturing real-world photographs for advanced AI training.
All images must be captured only through the official mobile application, following the
on-screen prompts exactly.
This is a precision-based project. Each task has strict instructions. Any deviation may lead to
rejection.
No prior professional photography experience is required, but attention to detail is mandatory.
General Rules (Applies to Every Image)
These rules apply across all categories and tasks:
● Capture real photos only using the in-app camera
● No AI-generated images
● No editing outside the app
● Use a tripod at all times
● The camera frame must remain completely stationary unless the prompt explicitly says
otherwise
● Ensure images are sharp, well-lit, and clearly visible
Consistency & Timing
● All images within a single task must be captured within a few minutes of each other
● Lighting must remain consistent across all images in the task
● Large time gaps or lighting changes may lead to automatic rejection during audit
Prompt Discipline
● Follow the on-screen prompt exactly
● If the prompt says “do not change anything else”, then absolutely nothing else should
change
● Do not improvise, substitute, or approximate objects, locations, or actions
Skipping Tasks
● If a task is not suitable for you, or if required objects or setups are unavailable, you may
skip the task using the in-app skip option
● Skipping tasks will not affect your account or future task allocation
Strictly avoid the following ❌:
● Logos, brand names, trademarks
● License plates or readable text
● Screens with content
● Offensive, illegal, unsafe, or sensitive content
If the prompt says “do not change anything else”, then absolutely nothing else should change.
Category A: Pair Editing Samples
You will capture 2 photos of the same scene.
General Requirements
● Each sample = 2 shots per scene (clear change, rest unchanged)
● Real-world indoor/outdoor/public environments (avoid cluttered or blank setups)
● Lighting consistent within each pair
● Camera stationary; same orientation & aspect ratio for both shots
● File format: JPEG, sRGB, size 4–50MB, aspect ratio 4:3 / 16:9 / 1:1
Sample Topics
● Position Change – Same subject moved to a different position
● Replacement – Swap one subject for another in the same spot
● Expression or Pose Change – Same model, same position, different expression or pose
Metadata
● Titles describe the change clearly
Example:
Shot 1: “soccer ball in foreground”
Shot 2: “change the position from foreground to background”
Restrictions
● Exclude IP - visible logos, brand names, license plates, and offensive content
Category C: Layered Dataset — Person with Objects
Start by capturing a full scene with 1 person and 1–3 objects in a natural, cohesive shot. Then
photograph each object separately as isolated shots.
1. Scene Composition
● Full scene shot:
○ One person + 1–3 objects in a natural setting.
○ Medium to wide framing; show person and surroundings.
○ Visually interesting composition with depth; avoid plain backgrounds.
2. Person Guidelines
● Maximum one person per set; diversity in age, body type, ethnicity, and abilities
encouraged.
● Same person can appear in up to 100 sets; same outfit allowed.
● Natural poses, looking at camera or not; avoid tight close-ups.
● Activities should feel real and active (e.g., cooking, hiking, biking).
3. Objects
● Person interacts naturally with 1–3 objects (e.g., laptop, coffee mug, backpack, pet
leash).
● Avoid cell phones and very small items (coins, keys, earbuds).
● Objects should be hand-sized or larger; clearly visible and recognizable.
● Object reuse allowed up to 5 sets; no reuse from previous submissions.
● Animals count as objects if interaction is clear (e.g., walking a dog).
4. Background
● Mix indoor (kitchen, living room) and outdoor (park, street) scenes equally.
● No moving elements in background; avoid logos, brands, or IP.
● Lighting: natural or artificial; create interesting shadows.
5. Isolated Shots
● Each object photographed alone in the same setting as the full scene.
● Same background, lighting, and surface whenever possible.
● Object fills ~20–40% of frame; maintain aspect ratio with main shots.
● Avoid extreme close-ups; crop only as needed for clarity.
6. Capture Sequence
1. Person with all objects.
2. Remove one object per shot (keep pose consistent).
3. Person only (no objects).
4. Empty scene (no person, no objects).
5. Isolated shots of each object.
Restrictions
● No AI-generated content.
● No Photoshop edits (adding/removing objects or models).
● Exclude IP: logos, brand names, license plates, offensive content.
Category D: Layered Dataset — Nested Objects
Start by capturing a full scene with 2–5 overlapping or touching objects arranged naturally. Then
remove one object at a time (starting with the closest to the camera) and photograph each step.
End with an empty scene and isolated shots of each
object in removal order.
1. Scene Composition
● Full scene shot:
○ 2–5 objects arranged so they overlap visually or touch physically.
○ Natural, realistic placement; all objects connected by a common surface (e.g.,
table, floor).
○ Ensure depth and clear layering for visual interest.
2. Object Guidelines
● Objects should vary in size and shape; avoid overly small items (coins, keys).
● Hero objects must be clearly visible, not cropped or blocked.
● Maintain physical connection (e.g., all objects touching the same surface).
● Removal order: start with the object closest to the camera, then proceed backward.
3. Background
● Use visually interesting indoor or outdoor settings with depth.
● Avoid plain backgrounds, moving elements, and trademarked logos or identifiable IP.
● Lighting: natural or artificial; consistent across all shots.
4. Capture Sequence
1. Full scene with all objects.
2. Remove one object at a time (closest first); capture each step.
3. Empty scene (no objects).
4. Isolated shots of each object in the same order as removed:
○ Same setting, background, and lighting whenever possible.
○ Object fills ~20–40% of frame; maintain aspect ratio.
Additional Requirement: Transparency & Reflections
● Same process as nested objects, but focus on:
○ Transparent or reflective surfaces (e.g., mirrors, glassware, candles, containers).
○ Ensure objects behind transparent items are visible and create realistic depth.
○ Capture natural reflections without distortion or artificial edits.
○ Lighting should highlight transparency and reflections without glare.
5. Restrictions
● No AI-generated content.
● No Photoshop edits (adding/removing objects).
● Exclude IP: logos, brand names, license plates, offensive content.
Category E: Character Consistency Task
Start by creating a narrative-based series of 100 photos featuring one main character across at
least 10 distinct settings and activities. Each photo must depict authentic, step-by-step actions
and be visually unique.
1. Mission Structure
● Dataset: One submission = 100 photos grouped into multiple scenes.
● Scenes:
○ 10 distinct scenes per dataset.
○ Each scene = ~10 shots showing related actions.
○ Scene IDs remain consistent within a scene; Shot IDs restart at 1 for each new
scene.
● Settings:
○ Mix indoor/outdoor and day/night environments.
○ Avoid domestic/home settings; include casual, professional, and public spaces.
2. Character Guidelines
● Main Character:
○ Age 30+; appears in every photo.
○ Fully visible (uncropped, unobscured) in at least 80% of shots.
○ Same character across all 100 photos; wardrobe changes across scenes.
○ Usage limit: up to 1000 total photos across all Missions.
● Secondary Characters:
○ 1–2 allowed; never shown without the main character.
● Casting:
○ Encourage diversity in race, ethnicity, gender, body size, age, and ability.
3. Shot Requirements
● Each photo must be distinct:
○ Vary pose, gesture, angle, focal length, emotion, and movement.
○ Avoid repetitive or staged shots.
● Composition:
○ Main character is primary focal point.
○ Use medium, full-body, and wide shots for variety.
● Lighting:
○ Prefer natural light; avoid harsh shadows or overexposure.
● Orientation:
○ Horizontal or vertical; remain consistent within a scene.
4. Storyline & Creative Direction
● Tell an authentic, progressive story:
○ Show clear beginning, middle, and end for each activity.
○ Include everyday tasks, routines, and interactions.
● Props & Wardrobe:
○ Blend personal style with modern and culturally unique details.
○ Avoid logos, brand names, and IP.
5. Metadata & Releases
● Metadata:
○ Required in English; include Scene IDs, Shot IDs, titles.
● Model Releases:
○ Required for all recognizable people; include demographic info (age, race,
gender).○
○ The artist name on release must match the contributor account.
● Incorrect metadata or missing releases = rejection.
6. Restrictions
● No AI-generated content.
● No Photoshop edits (adding/removing people or objects).
● No IP violations (logos, brands, trademarked designs).
● No offensive or prohibited content:
○ Nudity, profanity, sexually suggestive material.
○ Hate speech, violence, weapons, illegal activities.
○ Misleading or harmful content.
Common Reasons for Rejection
● Not following the prompt exactly
● Camera movement when not allowed
● Missing or incorrect images
● Logos or branded elements visible
● Blurry, dark, or unclear photos
● Repeating the same scene or shot
Final Reminder
The app already controls:
● Prompts
● Orientation
● Aspect ratio
● Task flow
Your responsibility is simple:
Follow the prompt exactly and capture clean, real images.
If you are unsure, do not guess. Stop and seek clarification by writing to us at SrotPix Helpdesk
(srotpix@[Link])
We also conduct regular webinars, you can join either of the two time slots (Monday-Friday):
Option 1
● 12:00 PM – 1:00 PM IST
● 06:30 AM – 07:30 AM UTC
Option 2 (Evening Session)
● 7:00 PM – 8:00 PM IST
● 13:30 PM – 14:30 PM UTC
Google Meet link (same for all sessions):
[Link]
Dial-in (optional):
● (US) +1 708-831-1650
● PIN: 476 470 362#
More phone numbers:
[Link]
We wish you the best!