Stalk-Net

STALK NET: A generic deep-learning based pipeline for stalk width-estimation. Faster-RCNN based region proposals are used for semantic segmentation of stalks. Stereo Imagery is used to get metric width.

Unsupervised Depth Estimation with GANs

Why We're Great >

UNSUPERVISED DEPTH ESTIMATION USING GANS: An Un-Supervised technique that exploits left-right consistency between pairs of stereo image to generate a depth map, that maps left image to right image.

Learning To Drive Using Images

LEARNING TO DRIVE USING IMAGES: An Environment Agnostic architecture that uses Variational Auto Encoders and Proximal Policy Optimization and Soft Actor Critic algorithms to learn policies that drive a car using semantically segmented monocular images.

Learning To Prune Grape Vines

LEARNING TO PRUNE GRAPE VINES: This project explores the use of Deep-RL for pruning grape-vines. A Neural Netwrok policy is trained using Proximal Policy Optimization to reach a pruning location in real time.

Robotic Stalk Grasping

ROBOTIC STALK GRASPING: A Deep-Learning based pipeline to detect grasp-points for taking Penetrometer readings in stereo images of Sorghum Stalks.

Autonomous Navigation in Vineyards

AUTONOMOUS NAVIGATION IN VINEYARDS: An Autonomous system that uses fusion of RTK GPS, Wheel Odometry and IMU data for localization. And uses a LIDAR for obstacle detection and avoidance.