4th Online Computer Vision and Artificial Intelligence Workshop (OnCV&AI)
August 30th – September 3rd, 2021, Tainan, Taiwan
We sincerely invite anyone in the world interested in spatial AI to participate. The workshop is held online and is free for everyone.
Sponsors
Program
Speakers
Live Stream
Recorded Talks
About
Our Sponsors
Time Zone
All times are listed in UTC+8, i.e. China Standard Time (CST) or Taipei Time, which is 8 hours ahead of Coordinated Universal Time (UTC).
Time Zone
All times are listed in UTC+8, i.e. China Standard Time (CST) or Taipei Time, which is 8 hours ahead of Coordinated Universal Time (UTC).
About the Workshop
This workshop is part of a series created for Ph.D. students, and other skilled engineers, to explore online computer vision and artificial intelligence techniques on the OAK-D camera. It consist of a few seminar speeches and a hackathon, distributed over a week.
During the hackathon everything is done online and implemented in real-time by our hackers on our OAK-D cameras as suggested by the audience. Everything is streamed on YouTube and Twitch.
Program
31
August
09:00 – 10:00
Alberto Speranzon,
Technical Fellow at Honeywell Aerospace
Computer Vision and Machine Learning for Aerial Autonomy
02
September
10:00 – 11:00
Jay Shah,
Arizona State University and ML Podcast host
Landscape of Explainable AI, Interpreting Deep Learning predictions and my observations from hosting an ML Podcast
03
September
14:00 – 15:00
Jose R. Chang,
National Cheng Kung University
Facial feature point detection using deep feature encodings
Subscribe to our workshop calendar
Speakers & Hackathon
Alberto Speranzon,
Technical Fellow at Honeywell Aerospace
Computer Vision and Machine Learning for Aerial Autonomy
August 31st, 09:00
Bio: Alberto Speranzon is a Technical Fellow within Honeywell Aerospace Advanced Technology. He received a Ph.D. in Electrical Engineering from the Royal Institute of Technology (KTH), Sweden in 2006. In September 2015 Alberto joined Honeywell where he is working on various aspects of autonomous systems for urban air mobility, leading such research areas as a program manager and principal investigator. Before joining Honeywell, he was a research scientist at United Technologies Research Center, CT (now Raytheon Technologies Research Center), where he covered various roles of increasing responsibility.
Alberto served as PI and Co-PI of various government-sponsored projects (DARPA, NASA) developing novel methods for navigation in GPS degraded/denied and tools to model multi-agent autonomous systems. Recently, he has been working on machine learning for advanced perception, modeling and verification of complex systems, and compositional abstractions for decision making. He is an IEEE Senior Member and a member of the Board of Governors of the IEEE Control Systems Society.
Abstract: In this talk I will present some past and current research activities related to post-acquisition image analytics and real-time video processing for a variety of applications centered around aerial vehicles. In particular, I will discuss image processing in the context of infrastructure inspection where images are collected by multi-rotor platforms as well as more recent work related to navigation and autonomous landing of unmanned aerial vehicles geared towards “air taxis” applications.
Jay Shah,
Arizona State University and ML Podcast host
Landscape of Explainable AI, Interpreting Deep Learning predictions and my observations from hosting an ML Podcast
September 2nd, 10:00
Bio: Jay is a 1st year Ph.D. student at Arizona State University supervised by Dr. Teresa Wu and Dr. Baoxin Li. Drawing upon the realms of biomedical informatics, computer vision, and deep learning, his research interests are in developing novel and Interpretable AI models for biomarker discovery and early detection of neurodegenerative diseases including Alzheimer’s and Post Traumatic Headache.
Prior to pursuing a Ph.D., he worked with Nobel Laureate Frank Wilczek, interned at NTU-Singapore, HackerRank-Banglore, and graduated from DAIICT, India. Jay also hosts an ML podcast (80,000+ downloads) where he has invited Professors, Scientists, Reporters, and Engineers working on different realms of research and applications of Machine Learning. More about him: public.asu.edu/~jgshah1/
Abstract: Using AI in application such as healthcare involve making crucial choices affecting human lives on a regular basis. However, research has shown that such judgments made by black-box models are inconsistent, error-prone, and most importantly difficult to interpret and trust. Hence explainable models are really important with the rising adoption of AI applications however the landscape of explainable and interpretable AI remains confusing and less explored. I will try and provide an overview of the landscape based on my research works and also from interviewing a record number of researchers, engineers, and professors working in similar domains on an ML podcast I host.
Jose R. Chang,
National Cheng Kung University
Facial feature point detection using deep feature encodings
September 3rd, 14:00
Bio: Jose Ramon Chang Irias obtained his M.Sc. in Mechanical Engineering (2017) and his two Bachelor’s degrees in Mechanical Engineering and Electronics Engineering (2016), all from Kun Shan University in Tainan, Taiwan.
Currently, he is a Ph.D. student in the Nordling Lab in National Cheng Kung University in Tainan, Taiwan. Where he is studying and researching Artificial Intelligence and Deep Learning. He has been a visiting researcher in the Center for Applied Intelligent Systems Research at Halmstad University in Halmstad, Sweden where he developed deep learning models to distinguish between the metabolic process of Dementia with Lewy Bodies (DLB) and Cognitively Normal (CN) subjects using the largest database for this disease in Europe collected by the European DLB consortium.
He is also a recipient of the Taiwan International Cooperation and Development Fund (ICDF) scholarship from 2012 to 2016, which is awarded to outstanding students from countries with diplomatic relationships with Taiwan. During his stay at NCKU he has also consecutively earned the Distinguished International Student Scholarship.
Abstract: Facial feature tracking is a key component of imaging ballistocardiography (BCG) where the displacement of facial keypoints needs to be accurately quantified to provide a good estimate of the heart rate. Traditional image registration algorithms can be divided into feature matching, like Scale Invariant Feature Transform (SIFT) or Speeded-Up Robust Features (SURF), and feature tracking, like the Lucas-Kanade method (LK). These have long represented the state-of-the-art in efficiency and accuracy. However, common deformations, like affine local transformations or illumination changes can cause them to fail.
Over the past five years, deep convolutional neural networks have outperformed traditional methods for most computer vision tasks. We propose a pipeline for feature tracking, in other words iteratively locating the same feature across all images in a video, that applies a convolutional stacked autoencoder to identify the most salient crops in an image and match them to a second image. The autoencoder learns to represent image crops into deep features encodings specific to the object category is trained on.
We compare the distance errors to thresholds defined at several significance levels in a Chi-square analysis of the differences in x and y coordinates from a series of relabeling attempts. When evaluated on manually labeled videos of faces, our method achieved a better mean error than all the other methods, except in one scenario where it had an insignificantly larger error of 0.31 pixels. More importantly, our method was the only one shown to not diverge under any of the testing circumstances, making it suitable for tracking applications.
These results show that our method creates better feature descriptors for image registration than the traditional algorithms.
Hackathon
September 3rd, 16:00 – 18:00
Interactive live exploration of spatial AI applications on the OAK-D camera based on suggestions from the speakers and audience. This is conducted by our hackers: Jose Ramon Chang, Jacob Chen, and Ric Tu, with Rain Wu handling communication with the audience.
Recorded Talks & Hackathon
Alberto Speranzon,
Technical Fellow at Honeywell Aerospace
Computer Vision and Machine Learning for the Aerial Autonomy
August 31st, 09:00
Jay Shah,
Arizona State University and ML Podcast host
Landscape of Explainable AI, Interpreting Deep Learning predictions and my observations from hosting an ML Podcast
September 2nd,
10:00
Presentation (files)
Jose R. Chang,
National Cheng Kung University
Facial feature point detection using deep feature encodings
September 3rd,
14:00
Hackathon,
Interactive live exploration of spatial AI applications on the OAK-D camera based on suggestions from the speakers and audience.
This is conducted by our hackers: Jose Ramon Chang, Jacob Chen, and Ric Tu, with Rain Wu handling communication with the audience.
September 3rd,
16:00
Past & Upcoming Workshops
About the Team
This workshop is arranged by the Nordling Lab located at the Dept. of Mechanical Engineering, National Cheng Kung University. It is sponsored by the National Cheng Kung University in Taiwan. We are part of the “National Cheng Kung University‘s Parkinson’s Disease Quantifiers” team lead by Ass. Prof. Torbjörn Nordling and Asc. Prof. Chi-Lun Lin selected as a finalist in the OpenCV AI competition 2021 out of more than 1400 teams globally.
The team consists of Dr. Akram Ashyani, Jose Chang, Esteban Roman, Tachyon Kuo, Gavin Vivaldy, Jacob Chen, Yushan Lin, Ric Tu, Prof. Chi-Lun Lin, and Prof. Torbjörn Nordling. We are working on quantification of motor degradation in Parkinson’s disease together with Prof. Chun-Hsiang Tan at Kaohsiung Medical University, and Dr. Chad Tsung-Lin Lee and Dr. Chung-Yao Chien at National Cheng Kung University Hospital. More precisely, on analysing the use of micro motions for assessment of motor degradation based on the Unified Parkinson’s Disease Rating Scale (UPDRS).
The National Cheng Kung University‘s Parkinson’s Disease Quantifiers team
About OAK-D
Luxonis and OpenCV AI Kit with Depth (OAK-D) has won the Best Camera and Sensor product of the year award by the Edge AI and Vision Alliance.
Luxonis is the company that has developed the OpenCV AI Kit. We received these cameras as finalists in the OpenCV AI competition 2021.
Making of the Workshop
About NCKU
The National Cheng Kung University (NCKU) is one of the two major comprehensive universities in Taiwan with app. 22 000 full-time students, attracting some of the best students in Taiwan. NCKU is number one in Academia-Industry Collaborations nationwide. Times Higher Education placed NCKU 38th in Impact Rankings 2020, 103rd in Asia University Rankings 2020. Academic Ranking of World Universities placed NCKU 301-400th internationally.
About OpenCV
OpenCV (Open Source Computer Vision Library) is an open source computer vision and machine learning software library. OpenCV was built to provide a common infrastructure for computer vision applications and to accelerate the use of machine perception in the commercial products. Being a BSD-licensed product, OpenCV makes it easy for businesses to utilize and modify the code.
The library has more than 2500 optimized algorithms, which includes a comprehensive set of both classic and state-of-the-art computer vision and machine learning algorithms. These algorithms can be used to detect and recognize faces, identify objects, classify human actions in videos, track camera movements, track moving objects, extract 3D models of objects, produce 3D point clouds from stereo cameras, stitch images together to produce a high resolution image of an entire scene, find similar images from an image database, remove red eyes from images taken using flash, follow eye movements, recognize scenery and establish markers to overlay it with augmented reality, etc. OpenCV has more than 47 thousand people of user community and estimated number of downloads exceeding 18 million. The library is used extensively in companies, research groups and by governmental bodies.
Along with well-established companies like Google, Yahoo, Microsoft, Intel, IBM, Sony, Honda, Toyota that employ the library, there are many startups such as Applied Minds, VideoSurf, and Zeitera, that make extensive use of OpenCV. OpenCV’s deployed uses span the range from stitching streetview images together, detecting intrusions in surveillance video in Israel, monitoring mine equipment in China, helping robots navigate and pick up objects at Willow Garage, detection of swimming pool drowning accidents in Europe, running interactive art in Spain and New York, checking runways for debris in Turkey, inspecting labels on products in factories around the world on to rapid face detection in Japan.
It has C++, Python, Java and MATLAB interfaces and supports Windows, Linux, Android and Mac OS. OpenCV leans mostly towards real-time vision applications and takes advantage of MMX and SSE instructions when available. A full-featured CUDAand OpenCL interfaces are being actively developed right now. There are over 500 algorithms and about 10 times as many functions that compose or support those algorithms. OpenCV is written natively in C++ and has a templated interface that works seamlessly with STL containers.
About AIA
Taiwan Artificial Intelligence Association (AIA) was founded in 2020, a leading purpose-driven organization for promoting commercialization of AI application, as well as advanced AI digital transformation for national and global industries.
About Crowdhelix
Crowdhelix is an Open Innovation platform that forges links between an international network of excellent researchers and innovating companies, so that they can plan and deliver pioneering collaborative projects Crowdhelix is open to applications from any organisation, of any size, anywhere in the world, that can demonstrate a strategic commitment to collaborative research and innovation. The network’s main focus is the European Union’s Horizon Europe programme, which has a €95.5 billion budget that funds thousands of collaborative research and innovation projects worldwide. Our goal is to connect leading research institutions and innovative companies around the world, so that together they can plan and deliver pioneering Horizon Europe projects.
About Luxonis
Our mission is to improve the engineering efficiency of embedding performant, spatial AI + CV into products. We do this by building and maintaining the open-source DepthAI ecosystem (more) which is also now the OpenCV AI Kit.
In other words, we build the core technology that allows human-like perception in real products – allowing 0-to-1 applications in nearly every industry. The technology allows solving problems that need human-like perception, but say in a 1″ cube. A good external writeup about Luxonis is on Bloomberg.
We are enabling what was science-fiction as of 2017:
- Wearable devices that perceive the world and allow the blind to perceive through soundscapes
- Embedded systems that can automatically protect endangered species
- Robots that enable organic farming with zero chemicals (by using lasers to target weeds and pests)
- Perception built into heavy machinery to real-time protect the health and safety of workers.
- Perception built into remote areas to autonomously monitor and protect the environment from leaks and other hazardous conditions.
- New forms of communication devices – bridging the gap between the in-person experience and the Zoom experience.
The above are just quick examples. There are countless solutions to fascinating problems being built off of our platform, with 15,000 devices shipped to date, in the hands of thousands of companies, some of which are working to integrate our technology into their own products.
We are a small startup with aims to have an outsized impact on the world. Still in the earliest stages, the solutions built with our technology have been covered in every top news, tech, and science publication in the world (Forbes, WSJ, Bloomberg, Washington Post, Engadget, etc.). We hold the records (independently) for the largest KickStarter raise for a computer vision project, for an AI project, and for a printed circuit board project.
Our mission is to materially improve the engineering efficiency of the world. This is what our team (and backers) previously did in another market (in business WiFi, with UniFi). And we now aim to do the same for this super-power of human-like perception in embedded systems – making it so prototyping with this power takes hours, and implementing it into a product is as fast as an enclosure can be designed and produced.
We are currently hiring:
Register Now!
We sincerely invite anyone in the world interested in spatial AI to participate. The workshop is held online and is free for everyone.
Organizer: The Nordling Lab @ National Cheng Kung University in Taiwan
Honorary Chair: Chair Prof. Cheng-Wen Wu
General Chair: Ass. Prof. Torbjörn Nordling
Technical Chair: Asc. Prof. Chi-Lun Lin
Assisting Chair: Akram Ashyani
Speaker Coordinators: Esteban Roman, Ray Chen
Lead Hacker: Jose Ramon Chang
Assisting Hackers: Jacob Chen, Ric Tu
Chat Host: Rain Wu
Streamer: Austin Su
Camera Operators: Ric Tu, YuShan Lin
Photographer: Gavin Vivaldy
Designer: Victor Osorio
Administrators: Winnie Tu, Anna Chu