Over 20 researcher from different country, I will list 3 research that I learned something in there (this is solely based on what I understand when participated the event, so there might be some mistake. Please refer to the researcher for more detailed info).
Towards More User Friendly Systems
Presentator：Akihiro Miyata from Nihon University
This research is about how the researcher think and make an user friendly system, especially for kids, elderly and disabled person. He thought of 3 solution for each case by using IoT.
Case one : IoT for Kids
Kids are not very skilled at searching web, therefore it takes times to find digital content.
Proposed:
IoB (Internet of Books), Kappan, document area identification method for extending book without markers. Allow to access digital area content by capturing area of book. This system consists of server and clients, OCR software to save area in database and send digital content to client.
Problem:
Difficult to obtain unique key from photo captured by kid.
Approach:
Utilize non reading direction as area specific key. This can yields the highest top 1 precision.
Case one : IoT for Elderly
Often forgot do important things, such as taking medicine, throw out thrash, etc. There is smart device can do it for them but they mostly they dislike it.
Proposed:
Internet of Furniture (maybe), xSeal, motion matching system. Use existing furniture to make it become smart. Develop by using accelerometers. Attach device and speaker to furniture. Registering and select desired function to furniture.
Problem:
Difficult to realize accurate furniture motion matching using single training data. Unrealistic to let elder make 100 training data.
Approach:
Most furniture are restricted in their motion. Utilizing finite axis pattern combination generated by the restricted furniture motion. Drawer type moves on 1 axis. Door type moves vertical axis. Mailbox type moves horizontal axis. System learn only axis pattern and is it occurred.
Case one : IoT for Disabled
Disabled people can't move freely in city owing to physical barriers.
Proposed:
BScanner, barrier detection method for making a barrier free . The goal is to make barrier free map and city planning. Two approachable method: Human judge and system judge.
Problem:
Human judge approach is highly accurate but cost much. System judge approach more accurate but lower coverage. Difficult to collect barrier information.
Approach:
Using people as sensor node. Use pedestrian acceleration data and deep learning.
Conclusion
How to make user friendly system:
- Have wide range of development skills.
- Always bear in mind that user only want to life comfortable and have compassion for users.
That is how inspiration come by thinking that two points.
What is Vection?
Presentator：Takeharu Seno from Kyushu University
This research mainly explain about vection. From what I understand, vection is about self motion perception. For example, when you ride a train and saw the train at beside move, you will feel like it is your own train that is moving. For more detailed info, open this link and video
Realizing Artificial Intelligence Concepts
Presentator：Kin Fun Li from University of Victoria
This research is mainly explain about AI and how and what did the researcher did with AI.
So, what is AI?
Based on my note from the presentation, to perform task normally require human intelligence, such as visual perception, speech recognition, decision making and translation between language. From researcher personal perspective, AI is search problem. Below is AI timeline (maybe):
- 1950: The Turing Test : evaluate question from human and computer, when evaluator don't know which answer come from
- 1960: HAL (Movie: "2001: Space Odyssey")
- 1970: MYCIN: Expert System
- 1980: Japan's FGCS
- 1990: Deep blue - World chess champion didn't won against IBM machine
- 2000: Movie: [AI] the movie
- 2010: Deep Mind - Computer system that can play game which is more complex game than chess at South Korea
Branch of AI:
- Machine learning
- Data Mining
- Deep Learning
Enabling tech:
- D-Wave quantum computer
- Intel nervana
- Google TPU (Tensor Processing Unit)
- Huawei Kirin 980 - AI tech inside, capable of analyzing thousand of images in second of time.