Making It Rain at ICLR 2018
Showcasing TwentyBN video recognition system to the global AI community
This spring, TwentyBN visited the scenic and rainy city of Vancouver to attend the International Conference on Learning Representations (ICLR) as a sponsor. In this post, we would like to share some highlights of ICLR and our technology, which we presented at our booth.
ICLR, chaired (among others) by Yoshua Bengio and Yann LeCun, is one of the most important annual conference for the Artificial Intelligence community. With over 2,000 attendees including researchers, corporate representatives, and students, ICLR is a crowded and vibrant bazaar, full of sparkling ideas and buzzes for machine learning. Naturally, the event offers a great venue for sharing and reviewing academic papers and technology demos.
Over the course of four days, leading researchers like Bernhard Schölkopf, Daphne Koller and Christopher Manning shared their knowledge to a large audience and pointed to new research directions. These talks were followed by intensive sessions of poster presentations and workshops among the attendees. Topics that popped up the most include GAN, deep reinforcement learning, and NLP. Many companies were also present, including Google, Amazon, Uber, and Element AI.
While the scientific progress presented at ICLR is exciting, not many companies at ICLR offered technology demos that bridge theories with the real world in an interactive manner. We were among the handful of companies that showcased a responsive AI technology that attendees could play with in real time.
This interactive demo we presented is Word Rain, a game that features our real-time dynamic gesture recognition system running on a CPU. The game will generate a random digit or letter that the player must draw in the air for the computer camera. If the camera recognizes that the player’s air gesture matches the randomly generated symbol, the game continues with a new random symbol; otherwise, game over. Word Rain was well received by many attendees, many of whom even invited their friends to play.
The popularity of our Word Rain demo at ICLR reflects our commitment to build powerful deep learning products that are interactive and practical for humans. With our scientific knowledge of AI ever increasing, it may be time for creators and innovators to build AI products that leave positive impacts on society. Indeed, we at TwentyBN are reaping the harvest of deep learning to build state-of-the-art technology that incorporates common sense, interacts with humans in real time, and improves our safety and life quality. Follow us to learn more about our progress!