GoodFuture AI Platform Wins UbiComp2020 Competition with Sensor‑Based Handwriting Recognition

Recently, the ACM International Conference on Pervasive and Ubiquitous Computing (UbiComp2020) announced its competition results, and GoodFuture's AI platform machine‑learning team stood out among more than 50 global teams to win the championship with a clear margin.

The UbiComp competition, co‑organized by the University of Erlangen‑Nuremberg’s Machine Learning and Data Analysis Lab and the Fraunhofer Institute for Integrated Circuits, attracted leading research groups from Google, Amazon, Microsoft, Alibaba, Tencent and many universities.

This year’s theme focused on frontier exploration in education, requiring participants to forgo traditional image‑based OCR and instead directly exploit smart‑pen trajectory data for character recognition. GoodFuture’s team employed an E‑RTCR model that integrates two cutting‑edge deep‑learning architectures—R‑Transformer and CRNN—to capture multimodal sensor signals such as stroke trajectory, angle, speed, acceleration, pauses, and pen‑lift events.

Compared with conventional OCR, this method overcomes challenges like highly abstract sensor signals, diverse writing habits, and large variations in multi‑sensor data distribution, resulting in scores far above the runner‑up.

Beyond the competition, GoodFuture has continuously invested in AI research, receiving Ministry of Science and Technology approval in August 2019 to build a national next‑generation AI open‑innovation platform for smart education. The team’s work has been featured in top conferences such as AAAI, WWW, AIED, and NCME, and they organized the first AI‑for‑Education workshop at AAAI.

GoodFuture now offers more than 100 AI capabilities for the education sector, with half being industry‑leading or unique models and weekly API calls exceeding 500 million, aiming to open these AI‑plus‑education solutions to the entire industry.