ECE Students Advised by Rama Chellappa win Best Poster Award and nVIDIA Best Paper Award at IEEE BTAS 2016

PhD students Upal Mahbub (ECE/UMIACS) and Sayantan Sarkar (ECE/UMIACS), and Assistant Professor Vishal M. Patel (ECE Rutgers, The State University of New Jersey), received “Best Poster” award at the IEEE 8^th International Conference on Biometrics: Theory, Applications, and Systems (BTAS 2016). The conference was held September 6 – 9^th, 2016, in Buffalo, New York.

Their paper, “Active User Authentication for Smartphones: A Challenge Data Set and Benchmark Results,” presents an investigation into automated user verification techniques for smartphones. The paper also introduces the Active Authentication Dataset 02 (AA02) for multi-modal user authentication research. Their research highlights the front camera, touch sensor, and location service features.

Benchmark results for face detection, face verification, touch-based user identification and location-based next-place prediction are also presented. The authors note that more robust methods that are adjusted specifically to the mobile platform are needed in order to achieve satisfactory verification accuracy.

In addition, PhD student Swami Swaminathan (UMIACS), Postdoc student Azadeh Alavi (UMIACS), and Research Associate, Carlos Castillo (UMIACS/CS), received the nVIDIA “Best Paper” award for their paper, “Triplet Probabilistic Embedding for Face Verification and Clustering.”

Their paper proposes an approach for solving unconstrained face verification problems that couples a deep CNN-based architecture with a low-dimensional discriminative embedding, and this was learned using triplet probability constraints in a large margin fashion.

The proposed pipeline enables faster training time and improves face verification performance especially at low false acceptance rate. In addition, it provides significant advantages in terms of memory and for post-processing operations such as subject specific clustering. The paper demonstrates the robustness of the resulting deep features to challenges including age, blur, pose and clutter by performing simple clustering experiments on recently released unconstrained datasets.

Swaminathan, Alavi, and Castillo were presented a Tesla K40 GPU for their paper.

BTAS 2016 is a premier research conference focused on all aspects of biometrics. It is intended to have a broad scope, including advances in fundamental signal processing, image processing, pattern recognition, and statistical and mathematical techniques relevant to biometrics.

Published September 20, 2016