Web Analytics Made Easy - Statcounter
Yihong Sun's Homepage

Yihong Sun

CS Ph.D. student at Cornell University
Contact: yihong-AT-cs-DOT-cornell-DOT-edu

About Me

I am a third-year CS PhD Student at Cornell University, advised by Prof. Bharath Hariharan. My work is supported by NSF GRFP and my research interests are computer vision and machine learning, especially in building visual perception systems that can learn from minimal supervision.

Previously, I obtained my Bachelor's degree from Johns Hopkins University where I studied Computer Science, Neuroscience, Applied Mathematics & Statistics, and Cognitive Science. During my undergradute studies, I worked with Bloomberg Distinguished Prof. Alan Yuille and Dr. Adam Kortylewski.

If you would like to chat with me, please drop me an email!

Publications

MOD-UV: Learning Mobile Object Detectors from Unlabeled Videos
Yihong Sun, Bharath Hariharan
ECCV 2024
We build an unsupervised mobile object detector from unlabeled videos only by leveraging independent motion information.
Dynamo-Depth: Fixing Unsupervised Depth Estimation for Dynamical Scenes
Yihong Sun, Bharath Hariharan
NeurIPS 2023
We improve unsupervised monocular depth estimation for dynamical scenes by modeling 3D independent flow and motion segmentation.
Amodal Segmentation through Out-of-Task and Out-of-Distribution Generalization with a Bayesian Model
CVPR 2022
We tackle amodal segmentation using a Bayesian generative model trained from non-occluded images and box-level annotations only.
Robust Instance Segmentation through Reasoning about Multi-Object Occlusion
CVPR 2021
We reason about multi-object self-occlusions by inspecting part-level activations of a Bayesian generative model.
Robust Object Detection Under Occlusion With Context-Aware CompositionalNets
CVPR 2020
(*) indicates joint first authors
We improve object detection under partial occlusion by regulating contextual bias and enhancing localization via compositional part voting.
Compositional Convolutional Neural Networks: A Robust and Interpretable Model for Object Recognition under Occlusion
IJCV 2020
We propose CompositionalNets, interpretable deep architectures with innate robustness to partial occlusion, for image classification and object detection.

Experience

Student Researcher
May 2024 - Now

Teaching

Graduate Teaching Assistant
Bowers CIS College of Computing and Information Science
- CS4670/5670 Introduction to Computer Vision (SP23)
- CS4787/5777 Principles of Large-Scale Machine Learning (FA22)
Undergraduate Course Assistant
Department of Computer Science
- EN.601.783 Vision as Bayesian Inference (SP22)
- AS.050.375/675 Probabilistic Models of the Visual Cortex (FA21, FA20)
- EN.601.226 Data Structures (SP21, FA20, SP20)