Naren Vasantakumaar

Successfully developed a pipeline to segment the object to a white background with an error rate of less than 8% using state-of-the-art segmentation model YOLOv8 on real-time images of excavator(object). •Successfully trained CycleGAN network to reproduce different textured synthetic images and produced a peak signal-to-noise ratio(PSNR) of 40, and Structural Similarity Index(SSIM) of 95% to the ground truth test data. •Successfully performed multi-object segmentation using YOLOv8

Pose estimation on crane using RGB images Keypoint and Joint angle detection using Computer vision and Deep learning algorithms Synthetic RGB image data generation using NVISII, Blender

Image segmentation and foreground seperation on synthetic images of an excavatorImage segmentation and binary classification on synthetic images of an excavator. Bounding box generation from RGB-D images

Worked on Motor control for designing Autonomous mobile manipulator of a Meccanum wheeled robot for localisation and control

Designed a control scheme for Quadcopter using PID and Fuzzy logic for trajectory tracking and Vertical take-off landing.