1st “Viscovery Paper Sharing Meetup” event was successfully held on Oct. 1st, 2016 at Viscovery’s Taipei office. The meetup gathered a number of Taiwan’s researchers in deep learning fields to share their latest research results. The participants included scholars from related fields. They had lively discussions with the presenters that day, and chose 5 outstanding papers that can get the 30,000 NTD reward.


Viscovery CEO Amos Huang hopes to encourage and subsidize Taiwan’s professors, researchers and students to apply and take part in top tier Computer Vision, Machine Learning, and Deep Learning conferences. We invited presenters who are accepted to ECCV, ACM MM, ICPR, and CVPR, these global conferences to share the most advanced technique and applications with Taiwan’s researchers, taking the preemptive opportunities in the new revolution of artificial intelligence (AI).


Visitors include professors and students from Academia Sinica (中研院), NTU (台大), NTHU (清大), NCKU (成大), NCCU (政大), CCU (中正), NTUST(台科大), NTUT(北科大), PCCU(文化) etc. The eight presenters (詹誠盛, 陳增鴻, 陳育聖, 陳威宇, 塗偉志, 楊惠芳, 吳依玲, 吳怡潔) shared their amazing research results on Computer Vision and Deep Learning that will be presented on international top tier conferences. NTHU professor Sun Min (孫民), who has close cooperation with Viscovery, also talked about their research with Stanford University, the algorithm of how to automatically recognize highlights of a video and generate a vivid description of it. The Academia Sinica researcher Y.-C. Frank Wang (王鈺強) also shared his research results of domain adaptation, “how to make machine learning more efficient”. In addition, the agenda included topics such as “algorithms to make panorama more lifelike”, “algorithms to automatically classify artwork”, “algorithms to detect objects more accurately”, and “how to make deep learning more efficient”. Due to information explosion and the help of GPUs, AI lets us deal with massive data in a short period of time, which can solve many different problems, and it has now become an indispensable technology in different industries. Viscovery’s core technique, FITAMOS, is based on deep learning. After recognizing photos, our technique can now successfully recognize video content, making it possible for machines to automatically analyze face, image, text, audio, motion, object, and scene in videos. With the support of our technique, we can systematically analyze every video, quickly find the appropriate entry point with business value for ads, and intelligently connect the plot with contextual ads, providing a better user experience. Viscovery has been cooperating with big video platforms such as Sohu, heading into the new video ad era with the latest video format VAST 4.0 released by IAB this June.


This event facilitated the exchange of views between the industry and the academics, helping Taiwan researchers to go faster and further in the deep learning field. We are looking forward to share Taiwan’s excellent research results with the world along with partners and experts in the AI and deep learning fields.

The following link leads to the title and abstract of the papers: