Purpose:
Image collection project, for the sole use of generating training data for machine learning and computer vision research. The image collection will be about collect images with Japanese text outdoors (like store fronts, street signs, billboards) and indoors
Requirements
• Job description: participant will see a prompt and they should upload a photo that matches this prompt (image with Japanese or Chinese text outdoors (like store fronts, street signs, billboards), when uplaoding, you need towrite a short description in english at the bottom, e.g. ´cheesecake store', just a few words, not a full sentence, then select the lighting conditions and upload the image.
Volume for each participant:
374 images for indoor
627 images for outdoor
Number of partipant we are hiring: 15 per language
Users need to be at least intermediate in English, as they need to read the guidelines that are going
to be in English and provide a short description in english of what is visible on the image they are uploading.
• Location: Based in Japan(for Japanese)
Based in Singapore(for Chinese).
Platform: Neevo
Data Validation & Acceptance Criteria
• Image uploaded must exactly align with the displayed prompt.
• Each participant must participate in the collection with one unique account.
• Each participant must have a full understanding of the project requirements and follow the job
instructions to ensure the most efficient throughput possible.
• No PII should be provided in any job within this project.
• Every image collected within the scope of this collection will be subject of a validation assessment
where Defined.AI will consider valid every image that has a positive agreement between multiple
validators.
• Users shouldn’t use the same items over an over, they should use a variety of items. Each image
should contain unique text
• Capture a variety of lighting conditions:
oLow - nighttime light, shadow, dimly lit interiors, obvious shadow
o Medium - indirect light
oHigh - bright sunlight, high artificial light
• Users will be asked to share images of items at a variety of distances:
oNear - textual object in hand or at an arm's distance
o Medium - textual object at 50 cm - 3 meters distance from you
oFar - textual object further than 3 meters away from you
• Image quality:
o minimum 720p
o not cropped, no flash, no filters/effects
o portrait (upright) orientation
Definition of invalid data:
Prompt/image mismatches.
Same account used by different participants.
User(s) participating with multiple accounts.
Image not following the guidelines requirements.
Image with another language instead of Japanese and Chinese.
3. Timeline and Deadlines
Start date: 19th February 2024
End date: 8th April 2024
Budget: $0.1 per image
This job is already closed and no longer accepting applicants, sorry.
