Meghana Basavaraj

Moved to Stuttgart, Germany and continuing working for Moovita Pte Ltd remotely as Consultant Working on BEVFusion for road segmentation using multi-camera images. Developed CUDA-based preprocessing in C++ to enhance execution time by 90% without compromising accuracy, involving CUDA kernel implementation and managing image data in GPU memory. Additionally, working on uncertainty estimation methods SNGP and DUE to improve handling of false predictions in deep learning models for object detection.

Built vision-based ADAS software on various embedded systems and developed lane detection and tracking modules. Ported the existing Lane departure warning system in python into C++. Execution time improved by 80%. Upgraded LDW system solution was deployed at 300 buses. Enhanced lane detection module for predicting additional attributes prediction (category, type and color), improved lane representation (Spline interpolation). Responsible for software integration, version control and releases.

Developed vision based analytics for an automatic fish biomass measuring tool. The analytics uses deep learning-based object and keypoint detection models with stereo-cameras. Achieved an average accuracy of ±5% relative to the actual weight. Winner of the "Most Innovative Team" at Siemens Hackathon 2019 held in Munich by demonstrating real-time object detection and tracking using Google TPU and Intel NCS2. Visited Munich, Germany for the event.

Developed Face tracker application for surveillance videos Developed Face recognition and attribute detection application (age, gender, name and pose) Worked on human pose estimation based analytics to detect safety gear in workers in a factory. Worked on development of camera streaming services and dockerization of analytics solutions.

Worked on Autosar compliant Device driver development of LIN, PORT, DIO, FlsLoader and IRQ for Aurix 2G Microcontrollers.These are bare metal drivers based on AUTOSAR (MCAL) and conform to ISO26262 Functional Safety standards. Worked on setting up a Unit Testing environment with Tessy Tool.