Tuesday, November 21, 2017
Neural Nets on the (Fashion) Runway
Researchers from the University of California San Diego and Adobe Research have demonstrated how artificial intelligence and neural networking could one day go beyond making recommendations about what fashion to buy, to helping retailers and apparel makers design bespoke clothing for consumers based on what they can learn from a buyer's preferences.
Their findings were published in early November in a paper* on "Visually-Aware Fashion Recommendation and Design with Generative Image Models." Computer Science and Engineering (CSE) professor Julian McAuley and his second-year Ph.D. student, Wang-Cheng Kang, teamed on the research with industry experts Chen Fang and Zhaowen Wang from Adobe Research.
"This suggests a new type of recommendation approach that can be used for recommendation, production and design," said CSE's McAuley and his colleagues. "These frameworks can lead to richer forms of recommendation, where content recommendation and content generation are more closely linked."
The project aimed to test how well tools from artificial intelligence (AI) and machine learning can help the fashion industry and consumers -- particularly those among the growing cohort of shoppers eager to buy clothing on the Internet.
recommendation is achieved by using
a 'visually aware' recommender based
on Siamese CNNs; generation is
achieved by using a Generative
Adversarial Net (GAN) to synthesize new clothing items in the user's
personal style. (Icons made by
Madebyoliver, Roundicons and Freepik
from www.flaticon.com )
While there are many algorithms and tools to help online retailers recommend designs to potential buyers, the UC San Diego-Adobe Research team went a giant step further. They wanted to see if it would be possible to crunch preference and other data not only to make recommendations, but potentially to enable computers to produce new clothing designs that would have an edge because they reflect a consumer's individual preferences.
Initially, the researchers focused on devising a system to create better recommendations, particularly in the case of 'visual' recommendations (where consumers can be swayed by how the product looks (as in the case of fashion apparel or artworks).
"Building effective recommender systems for domains like fashion is challenging due to the high level of subjectivity and the semantic complexity of the features involved," according to the new paper. They went on to note that visual recommendationcan be made more accurate by incorporating visual signals directly into the recommendation objective, using off-the-shelf feature representations derived from deep networks.
The team demonstrated that recommendation performance can be significantly improved by learning 'fashion aware' image representations directly, by training the image representation (from the pixel level) and the recommender system jointly. The paper grew out of recent work using Siamese Convolutional Neural Networks (Siamese-CNNs), so-called because they are a class of neural network architectures containing two or more identical subnetworks. (Siamese-CNNs are popularly used to find similarity or relationships between two comparable items.)
process for a given user and product category.
At left are real images; at right, synthetic ones.
The values are preference scores for each image.
For the fashion project, the researchers trained the Siamese-CNN to learn and classify a user's preferences for certain items.
From there, they used a neural networking framework called Generative Adversarial Network (GAN) to learn the distribution of fashion images and generate novel fashion items that maximize users' preferences. GANs train two networks on one set of data, and they have been particularly well-suited to generate realistic images. The resulting system can suggest items to buy from existing designs, but it can also be used to modify existing items, or to generate new designs tailored to a specific individual's preferences (based on 'big data ' about prior purchases, surveys, etc.).
"We show that our model can be used generatively, i.e., given a user and a product category, we can generate new images (in this case clothing items) that are most consistent with the user's personal taste," said first-author and CSE Ph.D. student Wang-Cheng Kang. "This represents a first step toward building systems that go beyond recommending existing items from a product corpus, to suggesting styles and helping to design new products.
The use of AI in the fashion industry is still in its infancy, but two of the world's largest online retailers -- Amazon and China's Alibaba -- are already working with AI tools, including GANs. As for the UC San Diego-Adobe Research project, the quality of the algorithmically-designed new clothes remains rudimentary at best. As CSE's McAuley told a writer from MIT Technology Review, "you'd have to read the tea leaves a little bit if you want to call that style or not." But, he warned, bringing neural networking to the fashion world is in its infancy.
______________________________
*Kang, W. C., Fang, C., Wang, Z., & McAuley, J. (2017). "Visually-Aware Fashion Recommendation and Design with Generative Image Models ", arXiv preprint, arXiv:1711.02231.
Cybersecurity Expert Accepts Test-of-Time Award for Introducing Return-Oriented Programming
At the 24th ACM Conference on Computer and Communications Security (CCS 2017) in early November, University of California San Diego professor Hovav Shacham was recognized for a paper he presented 10 years earlier that introduced the field of “return-oriented programming.”
left) accepts CCS Test-of-Time Award from
conference program chairs David Evans from
the University of Virginia (far left), Columbia’s
Tal Malkin, and Purdue’s Dongyan Xu (far right)
at CCS 2017 in Dallas.
The CCS Test-of-Time Award this year went to Shacham’s CCS 2007 paper for its lasting impact on security research and practice during the past decade. The CCS awards committee chose just one paper out of the 55 that appeared at CCS 2007, and Shacham was the sole author on that paper. Accordingly, the Computer Science professor is the only recipient of this year’s award.
Applications running on a computer keep track of what task they are performing and what tasks they are to perform next in a region of memory called the stack. Programming errors in these applications often allow the stack to be overwritten, confusing the application and causing it to misbehave or crash. Worse, an attacker who carefully crafts malicious input can confuse the application into running code he injects. Fixing all bugs in all applications is a hopeless task, so systems designers have devised defenses that rule out such code-injection attacks, by distinguishing “good” application code from attacker-introduced code. Since attackers must inject new code to achieve their ends, the thinking goes, ruling out the introduction of new code rules out attacks.”
Security researchers working in industry had developed techniques that allowed them to exploit buggy programs and cause them to undertake certain “bad” behavior even in the presence of defenses against code injection. Shacham’s paper* unified and generalized these exploit techniques under the rubric of what he jokingly called “return-oriented programming.” A return-oriented attack doesn’t introduce any new code. Instead, it makes use of short snippets of the original, “good” program’s code. The attacker combines these snippets in clever ways that allow him to synthesize any behavior he desires from them. "An analogy to return-oriented programming is a kidnapper who puts together a ransom note with letters cut from magazine headlines," said Shacham. "With all 26 letters gathered (and a photocopier), he can compose any message he wishes."
Using return-oriented programming, the attacker puts together a set of building blocks from which he can synthesize any desired behavior. (Such a set is said to be “Turing-complete.”)
Shacham’s techniques (and his jokey name for them) have stuck, both in academia and in industry. Shacham’s paper has been cited more than 900 times. Where his original work considered only Intel’s x86 family of processors, researchers have since extended Return-Oriented Programming (ROP) to other popular processors, including the ARM processors that power nearly all smartphones. A 2014 Microsoft report (see Microsoft Security Intelligence Report, Volume 16) found return-oriented techniques used in more than 90% of exploits targeting Microsoft products. Microsoft, Intel, and ARM have all announced new security mechanisms in their products specifically to mitigate the threat of return-oriented programming.
“The Test-of-Time Award is the most significant honor we bestow on a paper because it is not simply a reflection that a piece of research was strong and well-received, but that a decade later it has had significant impact on the field,” said CSE professor Stefan Savage, who directs the Center for Networked Systems (CNS), in which Shacham in a member. “Hovav's work on Return-Oriented Programming is an exemplar for such awards, because with one paper he made us all revisit our assumptions about what makes systems secure and the questions he posed in that work are still vibrant today. Moreover, this impact has not only been in academia, but Hovav's ideas have become a deep part of how real-world offense and defense is waged in computer security as well."
The 2017 conference took place in Dallas, TX, from October 30 to November 3. Shacham accepted the Test-of-Time Award during the conference’s banquet and awards ceremony on November 1. CCS is the flagship annual conference of the Special Interest Group on Security, Audit and Control (SIGSAC) of the Association for Computing Machinery (ACM).
Shacham did much of his research on the award-winning paper while a postdoctoral researcher at Israel’s Weizmann Institute of Science in 2006 and 2007, but he completed and presented the paper at CCS after joining the UC San Diego faculty in fall 2007.
That same year, he also participated in California’s “Top-to-Bottom” security review of the voting machines certified for use by the state’s then-Secretary of State, Debra Bowen. Shacham was part of the team reviewing Hart InterCivic source code, and the report he co-authored was cited by Bowen in her ultimate decision to withdraw approval for use of the Hart voting machines in California elections.
Shacham earned his Ph.D. in computer science in 2005 from Stanford University. His doctoral dissertation was runner-up for the Computer Science Department’s Arthur L. Samuel Thesis Award.
As an advisor, Shacham’s two most-recent Ph.D. graduates went to work at Google (Wilson Lian) in 2016 and Apple (Keaton Mowery) in 2015. In 2012, his student Stephen Checkoway joined Johns Hopkins University as a research professor, and more recently joined the faculty at the University of Illinois at Chicago.
______________________________________
*Shacham, Hovav, “The Geometry of Innocent Flesh on the Bone: Return-into-libc Without Function Calls (on the x86).” In Proceedings of the 14th ACM Conference on Computer and Communications Security, pp. 552-561. ACM, 2007 DOI: 10.1145/1315245.1315313
CCS 2007 Shacham Paper
CCS 2017 Awards
Former CSE Student Gets Funding for High-Flying Startup from Hollywood and the Pentagon
A small San Diego-based startup launched by a former student in CSE to reduce the threat of accidents due to civilian drones flying dangerously is getting some traction from an unusual place: Hollywood. The company is SkySafe , and the Hollywood star is Ashton Kutcher.
Ashton Kutcher and Air Force
Research Lab
In an exchange Nov. 15, the Air Force Research Lab used its Twitter account to pose a question. "Ashton Kutcher: What do you get when you mix drone experts from @MIT, @UCSanDiego and the @AFResearchLab?" Kutcher's reply Tweet: "An unprecedented level of airspace security. Check out @skysafeio for today's #techtuesday #investor".
So what is Kutcher's connection to SkySafe, and what does the company have in common with MIT, UC San Diego and the Air Force Research Lab?
Grant Jordan
The short answer to the latter question: SkySafe founder and CEO, Grant Jordan, a former Computer Science and Engineering M.S. student at UC San Diego. Jordan did his undergraduate engineering degree at MIT, then spent four years as a development engineer in the Air Force Research Lab testing anti-drone technology. From there he went to graduate school at UC San Diego, where he worked on cybersecurity issues in the lab of CSE professor Stefan Savage (recent winner of a 2017 MacArthur Foundation 'genius' grant).
Together with fellow CSE grad student, Paul Wicks (M.S. '14), Jordan created the cybersecurity-related startup SkySafe in 2015 and received $3 million in seed funding in 2016.
With the rapid growth in the use of civilian drones, SkySafe is developing technology to allow institutional users to disable drones flying in areas that are off-limits, or in areas that may be accessible but threatened by drones flying dangerously.
drone using RF technology developed by
San Diego-based SkySafe, co-founded by
former CSE students.
The technology leverages radio waves to override the instructions from a drone owner's remote-control unit, thus taking control of the airborne device. SkySafe uses radio frequency (RF) technology to send unwanted drones crashing to the ground. Alternatively, the technology can force a drone to return to its takeoff point automatically. Explained Jordan: "Our goal is to help drones be a positive change in society by improving safety, management and coordination."
So where does Ashton Kutcher fit into SkySafe's story? The actor co-founded the venture capital firm Sound Ventures in 2015, and the VC firm has invested in dozens of startups, most of them with founders under 30 years old, including SkySafe. In July, Kutcher's VC firm was one of eight companies that together invested $11.5 million of Series A funding in SkySafe. Other investors included the most powerful VC firm in Silicon Valley, Andreesen Horowitz, which had led the initial $3 million funding round for SkySafe in 2016.
In July, SkySafe also announced a $1.5 million contract with the Department of Defense. The contract is with the Pentagon's Silicon Valley office, known as the Defense Innovation Unit Experimental (DIUx), and Naval Special Warfare Command.
deployment of its RF counter-
drone system on light tactical
Polaris MRZR vehicle to prevent
drone attacks on Navy Seals in
the field.
The goal: to protect Navy Seals from small weaponized drones, primarily by reverse engineering communications and telemetry links which vary for each type or model of drone. DIUx aims to accelerate commercial innovation for national defense.
“It’s become a big problem on the military side,” SkySafe's Jordan told the news service FedScoop. “ISIS and other terrorist groups have started really weaponizing drones and bringing them into battles in a way that really hasn’t happened before.”
To counter unmanned aerial systems, SkySafe will install mobile RF counter-UAS systems on light tactical vehicles such as the Polaris MRZR (the military version of the off-road RZR vehicle) to protect Navy Seals from the new drone threat. SkySafe is currently working with Naval Special Warfare Command to do testing, and the startup expects to have a prototype ready for the field in 2018.
Finding a Faster, More Accurate Way to Identify Molecular Structures of Natural Products
Roughly 70 percent of drugs approved by the U.S. Food and Drug Administration are based on natural products such as plants and microorganisms in the soil or in the ocean. Now, four researchers from the Computer Science and Engineering department are part of an interdisciplinary team from UC San Diego that led development of a new method that is significantly faster and more accurate than existing methods for identifying the molecular structures of natural products.
Gary Cottrell
CSE professor Gary Cottrell is co-senior author on a paper* published online in the journal Nature Scientific Reports that spells out what the team calls Small Molecular Accurate Recognition Technology (SMART) and its benefits.
According to the paper’s authors, the new technique has the potential to achieve a ten-fold speed-up in the time it takes to identify a molecular structure from a natural product. As such, the SMART system could represent a new paradigm in chemical analysis and pharmaceutical drug discovery.
UC San Diego has a patent pending on the SMART technique. Named inventors on the patent application include both senior authors – CSE’s Cottrell and Scripps Institution of Oceanography professor William Gerwick – and both first authors: NanoEngineering Ph.D. student Chen Zhang (who works in Gerwick’s lab at Scripps), and CSE alumnus Yerlan Idelbayev (M.S. ’16), who worked on the project while still in CSE, but is now a Ph.D. student at UC Merced.
Technology (SMART).
Other CSE-affiliated co-authors of the paper include two undergraduate researchers from Prof. Cottrell’s lab who are majoring in Computer Science: junior Nicholas Roberts, who explored the effects of artificial experimental noise added to the SMART deep learning dataset; and senior Yashwanth Nannapaneni, who was a software engineering intern at Amazon over the summer. He expects to graduate next June.
In addition to the researchers from CSE, Scripps Institution of Oceanography and NanoEngineering, other UC San Diego co-authors of the article include faculty or students from the Skaggs School of Pharmacy and Pharmaceutical Sciences and the Department of Electrical and Computer Engineering. The team also includes individual researchers from Vanderbilt University, China’s Guangzhou Medical University and Germany’s University of Göttingen.
"The structure of a molecule is the enabling information," said UC San Diego oceanography and pharmaceutical sciences professor Bill Gerwick, Cottrell’s co-senior author on the new study. "You have to have the structure for any FDA approval. If you want to have intellectual property, you have to patent that structure. If you want to make analogs of that molecule, you need to know what the starting molecule is. It's a critical piece of information."
The SMART method uses a piece of spectral data unique to each molecule and then runs it through a deep learning neural network to place the unknown molecule in a cluster of molecules with similar structures. "The way we were able to accelerate the process is by essentially using facial recognition software to look at the key piece of information we obtain on the molecules," said Gerwick.
The key piece of information the team uses is called a heteronuclear singular quantum coherence (HSQC) nuclear magnetic resonance (NMR). Each HSQC NMR spectrum produces a topological map of spots that reveal which protons in the molecule are attached directly to which carbon atoms, an arrangement unique to each molecule.
CSE’s Cottrell and his team then developed a deep learning system that was trained with only around 2,000 2D images of HSQC spectra compiled from prior research. The convolutional neural network (CNN) took the images of spectra of unknown molecules and mapped them into a ten-dimensional space near molecules with similar traits.
“This is normally not enough data to train a deep network, but we used a technology called a Siamese network, in which you train on pairs of images,” said Cottrell. “This amplifies your training set by roughly the square of the number of compounds in a family, and is what made this project feasible.”
According to the article, as more compounds are added to the training set, “the SMART system will naturally improve in accuracy and robustness, thereby accelerating natural product structural elucidation and thus drug discovery.”
In their Nature Scientific Reports paper, the co-authors concluded that while they looked only at certain metadata associated with the spectra used in the study, “it is very possible to associate and integrate biological, pharmacological and ecological data with SMART, and thereby create new tools for enhanced discovery and development of biologically active natural products as well as other small molecules.”
____________________________________
*Chen Zhang, Yerlan Idelbayev, Nicholas Roberts, Yiwen Tao, Yashwanth Nannapaneni, Brendan M. Duggan, Jie Min, Eugene C. Lin, Erik C. Gerwick, Garrison W. Cottrell and William H. Gerwick, Small Molecule Accurate Recognition Technology (SMART) to Enhance Natural Products Research , Nature Scientific Reports, Issue 7, Article 14243, published online 27 October 2017. (doi: 10.1038/s41598-017-13923-x)
Related Links
SMART Article in Nature Scientific Reports
PhysOrg News Article
Cottrell Lab
Internet Society Honors CAIDA Director with Prestigious Service Award
CSE faculty-affiliate and alumna Kimberly (KC) Claffy (M.S. ‘91, Ph.D. ‘94) is the 2017 recipient of the prestigious Jonathan B. Postel Service Award. The Internet Society – a global non-profit dedicated to ensuring the open development, evolution and use of the Internet – called Claffy a “visionary in the field of Internet research.”
Claffy was selected by an international award committee comprised of former Postel Award winners. The committee places particular emphasis on candidates who have supported and enabled others in addition to their own contributions. In selecting Claffy for the honor, the committee cited her “pioneering work on Internet measurement through the development of infrastructure and methodologies for data collection, analysis, and sharing around the world.”
The Internet Society presented the award, including a $20,000 honorarium and a crystal engraved globe, on November 10 at the 100th meeting of the Internet Engineering Task Force (IETF), held in Singapore all this week.
San Diego's KC Claffy with Jonathan
B. Postel Service Award for her
"pioneering work on Internet
measurement." (Photo courtesy
Stonehouse Photographic/
Internet Society)
Claffy is the director and founder (in 1997) of the Center for Applied Internet Data Analysis (CAIDA), and a resident research scientist in the San Diego Supercomputer Center (SDSC) where CAIDA is based. Claffy is also a member of the Center for Networked Systems (CNS).
Her research interests span Internet topology, routing, security, economics, future Internet architectures and policy. Notably, Claffy is a pioneer in the field of measuring and understanding the Internet, not only through her research contributions, but her commitment to establishing and operating infrastructure to support large-scale data collection, curation, and sharing within the scientific research community.
Claffy was selected for the Postel Award by an international award committee comprised of former Postel Award winners. The committee placed particular emphasis on candidates who have supported and enabled others in addition to their own contributions.
The first of Claffy’s many papers on Internet traffic measurement and analysis was published in 1992, years before the Internet transitioned to the global, private sector-led network it is today. Since then, she has published dozens of papers and received numerous grants and awards for her work. In 1997, Claffy founded CAIDA as a center to conduct network research and build research infrastructure to handle large-scale data for the scientific research community.
“Simply put, Dr. Claffy’s long-standing and pioneering work has helped the global community better understand the Internet and how it is used,” explained Kathy Brown, President and CEO of the Internet Society, who presented the award. “In addition to leading the way in the field of Internet measurement and analysis itself, her dedication of resources to ensure widespread access to measurement data has allowed a range of disciplines -from network science and network operations to political science and public policy- to benefit from her efforts.”
In a profile of Claffy issued at the time of the award announcement, the scientist told the Internet Society that she hopes the past few decades of Internet development are miniscule when compared to where we are going. “I want to make the world safe for Internet science,” Claffy said. “The American people need better data – to understand what the Internet is, how it’s connected, and how data is being used.”
The Postel Award was established by the Internet Society to honor individuals or organizations that have made outstanding contributions to the data communications community. The award is focused on sustained and substantial technical contributions, service to the community, and leadership.
Jonathan B. Postel Service Award
Profile of KC Claffy
CAIDA
Kimberly Claffy Homepage
CSE Researchers Help Scripps Scientists Document and Model Coral Reefs for 100 Island Challenge
In a study published recently in the journal Coral Reefs, scientists from Scripps Institution of Oceanography created and analyzed detailed photomosaics of the coral reef at Palmyra Atoll in the Pacific Ocean. Led by Scripps coral-reef ecology Ph.D. student Clinton Edwards, the team canvassed more than 17,000 square feet of reef and 44,008 coral colonies, taking more than 39,000 images.
Palmyra Atoll,south of Hawaii.
(Photo by Stefani Gordon)
Rather than taking the conventional route of stitching together the collected images into high-resolution 2D image mosaics, Scripps researchers used cloud computing techniques developed by UC San Diego engineers and computer scientists to enable the creation of detailed models in 3D and even in 4D (i.e., adding the dimension of time). The models can be navigated and studied fully interactively, thanks to a team of technologists from UC San Diego’s Cultural Heritage Engineering Initiative (CHEI), led by UC San Diego structural engineering professor and computer science faculty-affiliate Falko Kuester.
5,000 corals; at bottom, identical photo
after being digitized by hand to label each
coral species with different color. 3D
technology can dramatically reduce the
time to capture coral reef data for later
analysis in the lab. Watch a video of
Palmyra Atoll coral reef annotated to
label species of coral in color overlays .
Creating the mosaics and wringing useful information out of them is a time-consuming process. Gathering the data, the scientists generally did three dives a day, and it took more than five full days of diving to collect the images for the 16 plots used in the study. Back in the lab, Edwards used a custom high-performance computing system to stitch together the 2,500-3,500 individual images that make up each mosaic. It takes the software several days to complete the rendering of the composite image, and around 100 hours to label and classify all the corals in each image. Then the final step is to extract the species information and analyze it, which takes another three days or so per image.
Digitization of the images is clearly the limiting step, said Edwards. But that may change soon, he added: “We have excellent collaborators in the Computer Science and Engineering department at UC San Diego, and are getting close to having a computer-assisted workflow that will dramatically accelerate this process.”
The data and scans of coral reefs from Palmyra Atoll represent some of the first results from the 100 Island Challenge (100IC), a project to create a global perspective on how coral reefs are changing over time. Scripps professor Stuart Sandin leads the overall project as principal investigator, with technology support from technologists in CHEI. The CHEI team is developing the novel 3D imaging and visual analytics techniques that Scripps scientists are using to create the mosaics that capture every detail of coral reef structure and ecology. So far, Scripps and international partners in 100IC have visited 70 of the 100 islands, and they plan to resurvey each site every two years. The resulting images, 3D models and analysis will also become baseline data to help local agencies study their own reefs.
related researchers (l-r): faculty-affilate Falko Kuester;
Ph.D. student Vid Petrovic; graduate student
Dominique Meyer; alumnus Chris McFarland,
and Computer Engineering alumnus Eric Lo.
In addition to Prof. Kuester, the 100 Island Challenge technology team from CHEI includes a handful of researchers affiliated with the CSE department: Ph.D. student Vid Petrovic, who is developing the software that the Scripps scientists use to create, visualize and analyze the coral reefs through 3D models and photomosaics; graduate student Dominique Meyer, who completed his B.S. in Physics at UC San Diego in 2016; software engineer and Computer Science alumnus Chris McFarland (B.S. '12) and Computer Engineering alumnus Eric Lo (B.S. '14), who is a staff robotics engineer in CHEI.
(The CHEI team has also deployed the technology for the Bermuda 100 Challenge , in which CHEI and Bermuda's Custodian of Historic Wrecks are creating digital replicas of shipwrecks and coral reefs surrounding Bermuda.)
reef at Palmyra Atoll.
CSE's Petrovic is now working on creating custom algorithms to accelerate the work on the coral-reef effort and other underwater mapping and computer modeling of the environment.
“More and more imagery is being collected across the field of marine sciences, and the pace and scale of the effort will only increase -- but more data doesn't automatically mean more, or better, science,” observed Petrovic. “It's an honor and a joy to be working so closely with a group of marine ecologists to address this, developing collaboratively the tools and workflows that are needed to make productive use of the imagery, whether for monitoring reef health, or for advancing basic science.”
According to Petrovic’s advisor, Prof. Kuester, the team is demonstrating that extremely detailed models consisting of billions of 3D data points can now be created overnight literally, when using distributed computing strategies. For a recent study, 14,000 images were turned into a comprehensive 3D model in less than a day. A digital surrogate of sorts, the model helped researchers annotate, measure and study the target environment in ways that would be impossible to achieve in-situ. By lifting the image data from 2D to 3D models or even 4D (including time), the technology is adding diagnostic value. Traditional analysis techniques when working on photomosaics can be supported by exporting a mosaic directly from the model, using the equivalent of an ultra-resolution virtual camera with user-definable characteristics.
to compile large-area photomosaics of
coral reef environments on which to
perform laboratory-based data
extraction.
Petrovic says the team is making it possible for scientists to virtually explore reefs in the lab, allowing them to time-travel from year to year and track the growth and decline of individual colonies, and to study spatial and temporal relationships across the reef.
“We're speeding up the digitization and annotation, and clearing a path to letting machine-learning techniques carry more of this burden,” Petrovic said. “This is all terribly exciting, and with much more to come. But the most rewarding aspect for me is the interdisciplinary collaboration that makes it possible in the first place, that lets us apply a decade of visualization research in support of vital ecological work.”
Beyond advancing the state of science in order to better safeguard coral reefs, the team’s research is literally giving coral reefs a voice. The 100 Island Challenge’s adoption of 3D digital tools allows the public at large to dive and experience these stunning ecosystems in virtual reality, to see what otherwise cannot be seen, and hopefully become stewards for coral reef preservation. “If a photo is worth a thousand words,” observed CHEI’s Kuester, “just imagine what a fully interactive 3D environment can accomplish for scientific storytelling.”
Read the original article on the Scripps Institution of Oceanography website.
Watch video of Scripps data gathering in Palmyra Atoll (YouTube).
Coral Reefs Journal Article
100 Island Challenge
Bermuda100 Challenge