My research focuses on using machine learning to improve sensor fusion, with a specific focus on multi-view matching. You can find my full research statement here
In this work we hope to combine the work from the robust optimization literature anddeep learning literature to create a general framework for robust matching. It can be usedto enhance performance in existing deep learning frameworks or to improve performancein robust learning frameworks. We leverage the aforementioned primal-dual trainingtechniques to learn more robust matching estimators. We formulate a Lagrangian primal-dual training framework for robust matching problems in asemi-supervised setting. Synthetic experiments have proved promising in a semi-supervised setting, and tests are underway for real-world data.
From my previous work, I have focused on computer vision with multiple cameras to obtain geometric information such as pose. More recent trends, such as in the self-driving car industry, have focused on incorporating data from multiple complementory sensors. This is called sensor fusion, and while more complex has many advantages over using a single kind of sensor. Cameras and radars complement each other’s information quite well, but research on fusing the two has only recently started to gain interest. Using graph neural networks, we hope to be able to use data-driven methods to fuse the sensors more robustly in more settings.
Published in IEEE Conference on Computer Vision and Pattern Recognition Workshop: Image Matching: Local Features and Beyond, 2019
Recommended citation: Stephen Phillips, Kostas Daniilidis, "All Graphs Lead to Rome: Learning Geometric and Cycle-Consistent Representations with Graph Convolutional Networks." IEEE Conference on Computer Vision and Pattern Recognition Workshop: Image Matching: Local Features and Beyond, 2019. https://arxiv.org/pdf/1901.02078.pdf
Published in Proceedings of International Conference on Learning Representations (ICLR), 2018
Recommended citation: Andrew Jaegle, Stephen Phillips, Daphne Ippolito, Kostas Daniilidis, "Understanding image motion with group representations." Proceedings of International Conference on Learning Representations (ICLR), 2018. https://arxiv.org/pdf/1612.00472
Quite different from my other work, this is project was a collaboration with the Frick Museum of Art in New York and UPenn. A fairly open ended project, our group explored various ways computer vision could complement art history. I supervising numerous projects including:
Master’s Thesis: Automatic Hierarchical Art Categorization with Few Shot Categories
Recommended citation: Susann Ludwig, Hongying Zhu, Stephen Phillips, Ashutosh Shiledar, Steve Feng, Derek Tseng, Leendert Ginkel, Michel Nielen, Aydogan Ozcan, "Cellphone-based detection platform for rbST biomarker analysis in milk extracts using a microsphere fluorescence immunoassay." In the journal of Analytical and bioanalytical chemistry, 2014. https://www.worldfoodinnovations.com/userfiles/documents/Article%20on%20Hormone%20Detection.pdf
Recommended citation: Isa Navruz, Ahmet Coskun, Justin Wong, Saqib Mohammad, Derek Tseng, Richie Nagi, Stephen Phillips, Aydogan Ozcan, "Smart-phone based computational microscopy using multi-frame contact imaging on a fiber-optic array." In the journal of Lab on a Chip, 2013. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3804724/
Recommended citation: Ahmet Coskun, Richie Nagi, Kayvon Sadeghi, Stephen Phillips, Aydogan Ozcan, "Albumin testing in urine using a smart-phone." In the journal of Lab on a Chip, 2013. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3810448/