Idea Details

Team and Contact Details

SNo Student Name School Degree Year Email
1Raja Zain AbbasSEECSUndergraduateFourth14beserabbas@seecs.edu.pk
1Hamza JavedSEECSUndergraduateFourth14besehjaved@seecs.edu.pk
1Syed Wahab Aftab ChishtySEECSUndergraduateFourth14beseschishty@seecs.edu.pk

Inter School Idea ? No
Do you need expertises from another area: No
If Yes please provide details of expertises you need:

Idea Details

Idea Name: Image Descriptor
Slogan: The Picture Speaks For Itself!
Supervisor Name: Dr Moazam Fraz
Supervisor Designation: Assistant Professor
Supervisor School: SEECS
Supervisor Department: Department of Computing
Contact number: 0000000000
Email ID: moazam.fraz@seecs.edu.pk
Abstract:
    Creating a autonomous system which is able analyze images and create detailed captions describing said image. The system is able to identify and describe relationships between core objects in image.
What is the unmet need in society that your idea will fulfill ?
    Humans receive around 80% information of environment through vision. So safe to say, computers which can see and make sense of images will have an essential role in the next technological revolution. It is estimated by 2022, computer vision market will grow to 48.6 billion, highlighting importance.
Who needs it ? How many would benefit ?
   The system will affect nearly every single field in the world. It can make repetitive tasks such as surveillance and tracking automatic. It can also provide assistance to visually impaired. It can also be used to extract conceptual meanings from images to be used in different applications, and more!
How will the solution works
    The basic process can be broken down into 4 parts. First, the input image goes through various image improvement techniques so no details are obscured. Then, the system identifies major regions in the images. These regions are then fed into another AI module which generates low level details of each region. Finally, these low level details are merged together with other details to form a single, coherent and logical sentence which describes the whole image and relationships between the regions.
Who are your competitors ? How is your solution different
    Similar implementations include Densecap and FasterRCNN. Our solution is different because we provide dense captioning. This involves not only labeling and describing regions in an image as our competitors do (such as densecap), but also relationships among regions and a overall description of image
Status: Approved
Entry Date & Time: 2017-12-15 (2309)