Ali Salman

Academic

Career

Verbatim text

BlindAssistant
Over the years as technology has progressed further and more and more smart solutions have 
emerged, assistive technologies all around the world also saw the light of innovation.
BlindAssistant is a solution aiming to cater the challenge of the ever growing technological 
advancements in assistive technologies using state-of-the-art techniques such as those in DL 
(FCHarDnet) and Computer Vision. Sometimes, the guide canes don’t offer their required safety 
levels because they don’t provide perception of the obstacles or objects types and also do not give 
information about the walking path. With that in mind, it shall leverage relevant information such 
as semantic segmentation map as well as depth map to output real-time feedback to the end user 
for various outdoor navigation scenarios. This required looking into various advanced models and 
techniques to prototype our main idea accordingly. We incorporated RGBD sensor Kinect XBOX v1 for 
obtaining 
depth 
feed 
in 
real-time.  
Features include: 
- Real-time terrain awareness 
- Obstacle avoidance system 
- Speech Engine for voice feedback 
- Flask server for application 
 
 
 
 
 
 
Technology Used: 
Python, Pytorch framework, Kaggle, Colab,  
Flask framework, LibFreenect (XBOX), Photoshop CC 
Supervisor Name: 
Dr. Asif Naeem 
Group Members:   
Ali Salman (17i - 0350)                    
M. Bilal Shabbir (17i - 0124)                    
M. Sohaib Akhtar (17i - 0330)

AI enrichment