Meghana Reddy Ganesina

I’m a second-year M.S. Computer Vision (MSCV) student at Robotics Institute, Carnegie Mellon University (CMU). I am working at CMU Argo AI Center for Autonomous Vehicle Research as a research collaborator, under the guidance of Prof. Deva Ramanan.

Recently, I worked at Zoox as a Prediction Machine Learning Intern. My work primarily involved designing an agent trajectory autoencoder inspired from Multi-Context Gating architecture for the downstream task of predicting consistent and diverse future agent trajectory set.

Email  |  CV  |  Google Scholar  |  Linkedin

I am most excited about developing robust and scalable computer vision algorithms. My research at Argo AI aims at developing an approach for LiDAR panoptic segmentation in an open-world setting. Previously, I have mainly worked on facial recognition systems and classification models for medical applications.

Developed a real-time disguised face identification system based on facial keypoint analysis using spatial Convolutional Neural Network. Furthermore, I collaborated with a team of researchers to annotate disguised face dataset to further the research.

Modelled a segmentation-free direct classification network to classify cervical cancerous cells which outperformed previous baselines. We observed correlations between the factors analyzed by pathologists and learned feature representations of the network.

This paper canvasses a strategy for Power-aware verification and discusses the steps to be followed in the run up to bring-up power aware simulations, scenarios to target for gate level with and without timing annotation, and optimizations for simulations that were done to achieve upto 52% higher performance and efficiency in the project execution.

Source code from Jon Barron