Manzareh
A significant proportion of the population is suffering from some or the other sort of vision impairment. The causes are many including cataract, Glaucoma, Diabetes or any physical accidents. Despite of loosing the biggest blessing of sight, the visually impaired community tries their best to make b
2025-06-28 16:34:04 - Adil Khan
Manzareh
Project Area of Specialization Artificial IntelligenceProject SummaryA significant proportion of the population is suffering from some or the other sort of vision impairment. The causes are many including cataract, Glaucoma, Diabetes or any physical accidents. Despite of loosing the biggest blessing of sight, the visually impaired community tries their best to make both ends meet on their own. However, they require some additional support for their living, which is most common in the form of any other person, or a stick.
This project focuses on resolving the difficulties for the visually impaired people, by making an app that reads the images for them. Using the concepts of deep learning and Artificial Intelligence it will
Project Objectives- To design a fully autonomous system for visually impaired that reduces the dependency on caretaker and guides them of their surroundings.
- A computer vision based algorithm that can deciphers the scene on a mobile application.
- To help blind people to get to know about the environment through a device.
- Complete software/application for a mobile device.
- Supervised learning algorithm for image captioning.
- To make a two-way communication device for visually impaired community.
uDeep Learning
uTensor Flow
uAndroid Studio
Benefits of the ProjectMaking the life of visually impaired easier
Technical Details of Final DeliverableMain Images or area of development that Manzarah will be targeting is Out Door activities. Such activities are all that are in good lighting condition. To be more specific, all such activities which are out side of a confined area or in a good natural light condition. More light means more corners and edges but that also means more possibilities.
Out Door activities also include a range of activities starting from beach to park to play ground to Roads, so my Application will target one type of these and when the model is ready for one type, we can iterate and repeat the same for all the other images. Sure, the conditions are different for different scenarios but it is the same, more or less.
Finally this will target object in general, included high priority towards object and scene recognition so a set of certain images containing known objects to the users will be fed for features extraction, this is the main purpose of Manzarah to be the actual Manzarah (Sight, Vision) of the Visually impaired people.
My project, as I described above, will need a set of images containing known objects to the user and the user will input captions of those images too, just to train the model specifically in accordance to that particular user. Different users can put in different objects of their own need.
So, the Application will now be using segmentation and lemmatization pin all possible words such as “My Car”, “My Keys” and “My Home” etc to the objects used by user in the training stage.
So my Text captions will look like, “My car is on the Road”, “My Keys are on the ground” etc
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 80000 | |||
| High Power Cuda Cluster | Equipment | 1 | 50000 | 50000 |
| Glasses With Camera and CPU | Equipment | 1 | 20000 | 20000 |
| blind people | Miscellaneous | 5 | 2000 | 10000 |