Co-Founder & CTO at Figr.design | Ex-Myntra, Adobe

Chirag Singla  • Peerlist

Various Image Processing techniques were optimized using Reinforcement Learning. In this, each pixel is considered as an agent and so locally it is decided that where to do what action.  

Image Enchancement Using Reinforcement Learning

An end to end video dubbing system which takes English videos as input and dubs them into various regional languages in real-time while producing human voice and conserving prosody of the original voice.

Auto Dub

It predicts the quality of seeds(good/bad) in real-time so that farmers can estimate the seeds before buying them so that they are not manipulated.

Seed Quality Prediction

Self-supervised model which auto-detects the drivable area for the wheelchair from depth image using V-disparity map. GitHub is home to over 50 million developers working together to host and review code, manage projects, and build software together. Self-supervised model which auto-detects the drivable area for the wheelchair from depth image using V-disparity map.

Road Anomaly Segmentation for Autonomous Wheelchair

Used Open pose and transfer learning method on different CNN layers to correct the user’s yoga pose in real-time and give feedback accordingly with 85% accuracy.

Yoga Pose Correction 

Automated attendance system, with zero to no human input required. Detects face through CCTV footage and marks attendance. Also, interaction and behavior of class were also analyzed to improve teaching quality.
 