Mark Hubenthal

Mark Hubenthal, Suren Kumar

Open-set logo recognition is commonly solved by first detecting possible logo regions and then matching the detected parts against an ever-evolving dataset of cropped logo images. The matching model, a metric learning problem, is especially challenging for logo recognition due to the mixture of text and symbols in logos. We propose two novel contributions to improve the matching model's performance: (a) using image-text paired samples for pre-training, and (b) an improved metric learning loss function. A standard paradigm of fine-tuning ImageNet pre-trained models fails to discover the text sensitivity necessary to solve the matching problem effectively. This work demonstrates the importance of pre-training on image-text pairs, which significantly improves the performance of a visual embedder trained for the logo retrieval task, especially for more text-dominant classes. We construct a composite public logo dataset combining LogoDet3K, OpenLogo, and FlickrLogos-47 deemed OpenLogoDet3K47. We show that the same vision backbone pre-trained on image-text data, when fine-tuned on OpenLogoDet3K47, achieves $98.6\%$ recall@1, significantly improving performance over pre-training on Imagenet1K ($97.6\%$). We generalize the ProxyNCA++ loss function to propose ProxyNCAHN++ which incorporates class-specific hard negative images. The proposed method sets new state-of-the-art on five public logo datasets considered, with a $3.5\%$ zero-shot recall@1 improvement on LogoDet3K test, $4\%$ on OpenLogo, $6.5\%$ on FlickrLogos-47, $6.2\%$ on Logos In The Wild, and $0.6\%$ on BelgaLogo.

Mark Hubenthal

The inverse source problem for the radiative transfer equation is considered, with partial data. Here we demonstrate numerical computation of the normal operator $X_{V}^{*}X_{V}$ where $X_{V}$ is the partial data solution operator to the radiative transfer equation. The numerical scheme is based in part on a forward solver designed by F. Monard and G. Bal. We will see that one can detect quite well the visible singularities of an internal optical source $f$ for generic anisotropic $k$ and $σ$, with or without noise added to the accessible data $X_{V}f$. In particular, we use a truncated Neumann series to estimate $X_{V}$ and $X_{V}^{*}$, which provides a good approximation of $X_{V}^{*}X_{V}$ with an error of higher Sobolev regularity. This paper provides a visual demonstration of the authors' previous work in recovering the microlocally visible singularities of an unknown source from partial data. We also give the theoretical analysis necessary to justify the smoothness of the remainder when approximating the normal operator.

Stephen Purpura, Dustin Hillard, Mark Hubenthal et al.

We present a system that enables rapid model experimentation for tera-scale machine learning with trillions of non-zero features, billions of training examples, and millions of parameters. Our contribution to the literature is a new method (SA L-BFGS) for changing batch L-BFGS to perform in near real-time by using statistical tools to balance the contributions of previous weights, old training examples, and new training examples to achieve fast convergence with few iterations. The result is, to our knowledge, the most scalable and flexible linear learning system reported in the literature, beating standard practice with the current best system (Vowpal Wabbit and AllReduce). Using the KDD Cup 2012 data set from Tencent, Inc. we provide experimental results to verify the performance of this method.