Deepti Hegde

I am a PhD student advised by Dr. Vishal Patel at the Vision and Image Understanding Lab at Johns Hopkins University where I work on 3D computer vision and deep learning. I am interested in learning in limited-label scenarios and vision-language models for 3D scene understanding.

I spent a lovely summer as an intern at Mitsubisihi Electric Research Labs (MERL), where I worked with Suhas Lohit, Kuan-Chuan Peng, and Mike Jones on domain generalization for 3D object detection.

I did my Bachelor' degree in Electronics and Communication at KLE Technological University, India where I was advised by Dr. Uma Mudenagudi and worked on pointcloud refinement, image enhancement, and embedded intelligence.

dhegde1[at]jhu[dot]edu  /  CV  /  Google Scholar  /  Twitter  /  Github

I am interested in computer vision, deep learning, and 3D pointcloud processing. I am always open to participating in collaborations! Feel free to reach out if you want to work with me on problems in 3D computer vision and scene understanding.

CLIP is not suitable for extracting 3D geometric features as it was trained on only images and text by natural language supervision. We work on addressing this limitation and propose a new framework termed CG3D (CLIP Goes 3D) where a 3D encoder is learned to exhibit zero-shot capabilities.

Addressing the limitations of traditional feature aggregation methods for prototype computation in the presence of noisy labels, we utilize a transformer module to identify outlier ROI's that correspond to incorrect, over-confident annotations, and compute an attentive class prototype. Under an iterative training strategy, the losses associated with noisy pseudo labels are down-weighed and thus refined in the process of self-training.

In order to avoid reinforcing errors caused by label noise, we propose an uncertainty-aware mean teacher framework which implicitly filters incorrect pseudo-labels during training. Leveraging model uncertainty allows the mean teacher network to perform implicit filtering by down-weighing losses corresponding uncertain pseudo-labels.

We propose a physics-based approach to simulate lidar point clouds of scenes in adverse weather conditions. These augmented datasets can then be used to train lidar-based detectors to improve their all-weather reliability. Specifically, we introduce a hybrid Monte-Carlo based approach that treats (i) the effects of large particles by placing them randomly and comparing their back reflected power against the target, and (ii) attenuation effects on average through calculation of scattering efficiencies from the Mie theory and particle size distributions.

We propose a method to refine sparse point clouds of complex structures generated by Structure from Motion in order to achieve improved visual fidelity of ancient Indian heritage sites.

source code