Samsung Electronics’ Global Research & Development (R&D) Centers have presented their studies to the CVPR (Computer Vision and Pattern Recognition) introducing new computer vision, deep learning and AI related technical researches.

CVPR is the world’s biggest conference on computer engineering and AI. At this year’s conference, held online from June 14 to 19, Samsung Research, an advanced R&D hub within Samsung Electronics’ SET Business and its advanced R&D centers gave presentations on a total of 11 thesis papers. Researchers from Samsung Moscow AI center and Samsung Toronto AI center were invited to oral presentations, an opportunity given to only 5% of the entire attendees.

At the oral presentation, Pavel Solovev of Samsung Moscow AI Center introduced ‘High Resolution Daytime Translation without Domain Labels’, which is a technology that changes a high resolution landscape photograph into scenes from various times of the day using data without domain label. Konstantin Sofiiuk also introduced ‘f-BRS: Rethinking Backpropagating Refinement for Interactive Segmentation’, which is a technology that allows a user to simply click an object in a photograph to precisely select and separate it.

‘High Resolution Daytime Translation without Domain Labels’

Joining from the Toronto AI Center, researcher Michael Brown and his team introduced the paper titled ‘Deep White-Balance Editing’, which was also selected for an oral presentation. This AI technology corrects white-balance mistakes made in a captured photograph much more accurately than existing photo editing programs. This technology also allows users to accurately adjust the photo’s white-balance color temperature.

Deep White-Balance Editing

Researchers from Samsung Research America also presented interesting findings at the conference. Eric Luo’s study titled ‘Wavelet Synthesis Net: An Efficient Architecture for Disparity Estimation to Synthesize DSLR Calibre Bokeh Effect on Smartphones’ focused on key enablers to narrow the gap between DSLR and smartphone camera in terms of bokeh, the narrow depth of field (DoF).

Yilin Shen from Samsung Research America’s AI Center introduced a study on out-of-distribution (OoD) benchmarks for deep neural networks research. Shen’s study titled ‘Learning out-of-Distribution Detection without Out-of-Distribution Data’ proposed the key machine learning algorithm of drastically improving the detection rate, one of major challenges in AI technology.

Additionally, the studies proposed by researchers from the Samsung Research’s Visual Technology team and Samsung R&D Institute India-Bangalore were also selected by CVPR.

Pavel Ostyakov, one of the researchers from SAIC-Moscow, won first place among 110 teams at the “Inclusive Images Challenge,” a Kaggle¹ competition hosted by Neural Information Processing Systems (NeurIPS) 2018, which took place in Montreal, Canada from December 3^rd to the 8^th. NeurIPS, formerly known as NIPS, is the world’s largest conference in the field of AI. As of 2017, a total of 8,000 people participated at the event. Apart from machine learning and neuroscience, experts from many related research fields such as cognitive science, computer vision, statistical linguistics, and information theory actively participate in the conference. Winning this challenge was a significant achievement for both Samsung and Pavel as he has been named the Competitions Grandmaster in the Kaggle category of data science expertise, which is the highest tier possible. Pavel also has the honor of being ranked one of the world’s top five scientists on Kaggle.

In the “Inclusive Images Challenge”, participants developed image recognition and classifier models that can successfully perform image classification tasks even when the test images are geographically and culturally different from the previously shown images in which the recognition models were trained for.

(Left image, from left) Pavel Ostyakov, a Researcher at Samsung AI Center-Moscow (SAIC-Moscow), and Jin Wook Lee, the Head of Samsung R&D Institute Russia. (Right image) A snapshot of SAIC-Moscow’s opening ceremony, which took place in May of this year

For example, an image classifier may fail to properly apply “wedding” related labels, such as “bride,” “groom,” and “celebration,” to an image, if a couple is not wearing traditional western European wedding attires or colors. This challenge attempts to address the biases that exist in many of the most popular training datasets. Through this challenge, researchers can identify ways to teach image classifiers by generalizing the accumulated data and apply them in new geographic and cultural contexts. The expectation is that the scientific community will make even more progress in inclusive machine learning that benefits everyone, everywhere.

Samsung has been a strong proponent for inclusive and fair AI in which the two principles are deeply incorporated into its AI developments, emphasizing the need for a truly global and bias-free artificial intelligence as it plays a bigger role in society. Samsung’s joining of the Partnership on AI (PAI) in November is also a part of this effort.

Last September, the SAIC-Moscow team also participated in “The 2nd YouTube-8M Video Understanding Challenge,” hosted by the European Conference on Computer Vision (ECCV) 2018. ECCV is one of the world’s top research conferences in the computer vision area and is held biennially. In this Kaggle competition, researchers were provided with public YouTube-8M training and validation datasets that consists of millions of videos with labels, and then asked to develop classification algorithms, which accurately predicts the labels of 700,000 previously unseen YouTube videos. During the competition, the SAIC-Moscow team utilized a unique approach in its complex model and data analysis, placing second by a very narrow margin.

An exterior view of the White Square Business Center, where SAIC-Moscow is located

“We love these competitions because they provide us with opportunities to measure ourselves to participate with the best in the AI industry,” said Pavel Ostyakov at SAIC-Moscow. “Researchers at Samsung are obsessed with making AI a part of everyday lives. So, it is exciting to take part in the challenges where we can contribute our skills to develop the technology.”

The global AI centers reflect Samsung’s commitment to next generation AI development. Besides Russia, there are six others located across the globe – Korea, Silicon Valley and New York in the U.S., Toronto and Montreal in Canada, and Cambridge, the U.K. Each location is focused on a different area of strengths and leverages its unique characteristics. These AI centers are playing a pivotal role in implementing Samsung’s vision for human-centric AI technologies and products.

* 1 An online community allowing users to find and publish datasets, explore and build models in a web-based data-science environment, work with others, and enter competitions to solve data science challenges.