Saksham Chaurasia

Collaborated in the research & development of an end-to-end pipeline for detecting & tracking objects like person/pet in real-time and ensuring their safety through efficient distress detection around the pools. Reduced 15% execution time from development to production by defining the roadmap, and structured pipeline from data processing to deployment. Designed and implemented a sophisticated tracking algorithm that achieved state-of-the-art results on the MOT benchmark with an accuracy improvement of almost 30% in detection.

Led the development of an end-to-end pipeline for efficiently managing the assembling & kitting process for large warehouses & factories using computer vision techniques. Implemented & deployed efficient deep learning models for object detection, tracking, pose estimation, and activity recognition from videos.

Led the research & development for building robust Identity Management algorithms such as Face Match, OCR, and Anti-Forgery, liveliness, etc. Developed and achieved state-of-the-art results on the Face Match algorithm with 99.38% accuracy on the Faces in the Wild dataset and a response time of less than 100ms. Designed & implemented a highly efficient OCR algorithm from scratch with an error rate of less than 5%.

Contributed in research & development of an efficient algorithm for Breast Tumor Detection, Classification, and Localization. Developed a new deep learning segmentation network 'GRA U-Net' which clocked an accuracy of 99% and F1 measure of 92%.

Project/Thesis: Plant Disease Identification using Convolutional Neural Network