Image Collection
Budget: ₹75,000 – ₹150,000 INR
500 Images per object listed above = 2,58,500 Images
Types of objects (colour, shape, size and texture)
Types of materials
Materialistic objects have high variation to cover all possible scenarios related to material. Non materialistic objects will be covered in natural form.
Colour
Objects to be covered with equal distribution among possible colours
Angle Variation
A scene can be captured in maximum 4 directions for large objects such as BUS / Truck for small object only one directions image will get approve
Height Variation
The height from which the object is captured should not be very high and the object should be clearly captured.
Distance Variation
The distance should be enough that the object is easily identifiable.
Diversity
The object scene should be realistic and more than one object per frame.
Lighting Variation
The images captured should be uniformly distributed in the natural good lighting conditions and low lighting conditions. The prime constraint is only that the object should be visible and recognizable.
Deformation and Occlusion
Deformed and occluded images will be accepted provided the number of images per object is less than 5%
Image Resolution
● Minimum 640*480,
● 720p & 1080 p recommended
● Camera resolution 10mp
Camera
● Mobile
Orientation
● Portrait -20%
● Landscape -80%
Aspect Ratio
4:3, 3:4, 16:9, 9:16 and 1:1
Acceptance criteria for collected data
All Images/scenes should be unique.
Selection Criteria:
30% of the batch to be checked, 95% of the selected to pass for the batch to be considered pass.
GPS variance:
GPS data for all images to be captured. The images to be equally distributed across the country.
Exclusions
● Zoom Images
● Fabricated images
● Web scrapped images
● Images scrapped from Video
Image with watermarks
Types of objects (colour, shape, size and texture)
Types of materials
Materialistic objects have high variation to cover all possible scenarios related to material. Non materialistic objects will be covered in natural form.
Colour
Objects to be covered with equal distribution among possible colours
Angle Variation
A scene can be captured in maximum 4 directions for large objects such as BUS / Truck for small object only one directions image will get approve
Height Variation
The height from which the object is captured should not be very high and the object should be clearly captured.
Distance Variation
The distance should be enough that the object is easily identifiable.
Diversity
The object scene should be realistic and more than one object per frame.
Lighting Variation
The images captured should be uniformly distributed in the natural good lighting conditions and low lighting conditions. The prime constraint is only that the object should be visible and recognizable.
Deformation and Occlusion
Deformed and occluded images will be accepted provided the number of images per object is less than 5%
Image Resolution
● Minimum 640*480,
● 720p & 1080 p recommended
● Camera resolution 10mp
Camera
● Mobile
Orientation
● Portrait -20%
● Landscape -80%
Aspect Ratio
4:3, 3:4, 16:9, 9:16 and 1:1
Acceptance criteria for collected data
All Images/scenes should be unique.
Selection Criteria:
30% of the batch to be checked, 95% of the selected to pass for the batch to be considered pass.
GPS variance:
GPS data for all images to be captured. The images to be equally distributed across the country.
Exclusions
● Zoom Images
● Fabricated images
● Web scrapped images
● Images scrapped from Video
Image with watermarks