Uses Archives - Business Intelligence Info

Tag Archives: Uses

Google’s Objectron uses AI to track 3D objects in 2D video

March 12, 2020 Big Data

Coinciding with the kickoff of the 2020 TensorFlow Developer Summit, Google today published a pipeline — Objectron — that spots objects in 2D images and estimates their poses and sizes through an AI model. The company says it has implications for robotics, self-driving vehicles, image retrieval, and augmented reality — for instance, it could help a factory floor robot avoid obstacles in real time.

Tracking 3D objects is a tricky prospect, particularly when dealing with limited compute resources (like a smartphone system-on-chip). And it becomes tougher when the only imagery (usually video) available is 2D due to a lack of data and a diversity of appearances and shapes of objects.

The Google team behind Objectron, then, developed a toolset that allowed annotators to label 3D bounding boxes (i.e., rectangular borders) for objects using a split-screen view to display 2D video frames. 3D bounding boxes were overlaid atop it alongside point clouds, camera positions, and detected planes. Annotators drew 3D bounding boxes in the 3D view and verified their locations by reviewing the projections in 2D video frames, and for static objects, they only had to annotate the target object in a single frame. The tool propagated the object’s location to all frames using ground truth camera pose information from AR session data.

To supplement the real-world data in order to boost the accuracy of the AI model’s predictions, the team developed an engine that placed virtual objects into scenes containing AR session data. This allowed for the use of camera poses, detected planar surfaces, and estimated lighting to generate physically probable placements with lighting that matches the scene, which resulted in high-quality synthetic data with rendered objects that respected the scene geometry and fit seamlessly into real backgrounds. In validation tests, accuracy increased by about 10% with the synthetic data.

Better still, the team says the current version of the Objectron model is lightweight enough to run in real time on flagship mobile devices. With the Adreno 650 mobile graphics chip found in phones like the LG V60 ThinQ, Samsung Galaxy S20+, and Sony Xperia 1 II, it’s able to process around 26 frames per second.

The Objectron is available in MediaPipe, a framework for building cross-platform AI pipelines consisting of fast inference and media processing (like video decoding). Models trained to recognize shoes and chairs are available, as well as an end-to-end demo app.

The team says that in the future, it plans to share additional solutions with the research and development community to stimulate new use cases, applications, and research efforts. Additionally, it intends to scale the Objectron model to more categories of objects and further improve its on-device performance.

Let’s block ads! (Why?)

Big Data – VentureBeat

GitHub now uses AI to recommend open issues in project repositories

January 22, 2020 Big Data

Large open source projects on GitHub have intimidatingly long lists of problems that require addressing. To make it easier to spot the most pressing, GitHub recently introduced the “good first issues” feature, which matches contributors with issues that are likely to fit their interests. The initial version, which launched in May 2019, surfaced recommendations based on labels applied to issues by project maintainers. But an updated release shipped last month incorporates an AI algorithm that GitHub claims surfaces issues in about 70% of repositories recommended to users.

GitHub notes that it’s the first deep-learning-enabled product to launch on Github.com.

According to GitHub senior machine learning engineer Tiferet Gazit, GitHub last year conducted an analysis and manual curation to create a list of 300 label names used by popular open source repositories. (All were synonyms for either “good first issue” or “documentation,” like “beginner friendly,” “easy bug fix,” and “low-hanging-fruit.”) But relying on these meant that only about 40% of the recommended repositories had issues that could be surfaced. Plus, it left project maintainers with the burden of triaging and labeling issues themselves.

The new AI recommender system is largely automatic, by contrast. But building it required crafting an annotated training set of hundreds of thousands of samples.

GitHub began with issues that had any of the roughly 300 labels in the curated list, which it supplemented with a few sets of issues that were also likely to be beginner-friendly. (This included those that were closed by a user who had never previously contributed to the repository, as well as issues closed that touched only a few lines of code in a single file.) After detecting and removing near-duplicate issues, several training, validation, and test sets were separated across repositories to prevent data leakage from similar content, and GitHub trained the AI system using only preprocessed and denoised issue titles and bodies to ensure it detected good issues as soon as they’re opened.

In production, each issue for which the AI algorithm predicts a probability above the required threshold is slated for recommendation, with a confidence score equal to its predicted probability. Open issues from non-archived public repositories that have at least one of the labels from the curated label list are given a confidence score based on the relevance of their labels, with synonyms of “good first issue” awarded higher confidence than synonyms of “documentation.” At the repository level, all detected issues are ranked primarily based on their confidence score (though label-based detections are generally given higher confidence than ML-based detections), along with a penalty on issue age.

Data acquisition, training, and inference pipelines run daily, according to Gazit, using scheduled workflows to ensure the results remain “fresh” and “relevant.” In the future, GitHub intends to add better signals to its repository recommendations and a mechanism for maintainers and triagers to approve or remove AI-based recommendations in their repositories. And it plans to extend issue recommendations to offer personalized suggestions on next issues to tackle for anyone who has already made contributions to a project.

Let’s block ads! (Why?)

Big Data – VentureBeat

Ubisoft uses AI to teach a car to drive itself in a racing game

December 29, 2019 Big Data

Reinforcement learning, an AI training technique that employs rewards to drive software policies toward goals, has been applied successfully to domains from industrial robotics to drug discovery. But while firms including OpenAI and Alphabet’s DeepMind have investigated its efficacy in video games like Dota 2, Quake III Arena, and StarCraft 2, few to date have studied its use under constraints like those encountered in the game industry.

That’s presumably why Ubisoft La Forge, game developer Ubisoft’s eponymous prototyping space, proposed in a recent paper an algorithm that’s able to handle discrete, continuous video game actions in a “principled” and predictable way. They set it loose on a “commercial game” (likely The Crew or The Crew 2, though neither is explicitly mentioned) and report that it’s competitive with state-of-the-art benchmark tasks.

“Reinforcement Learning applications in video games have recently seen massive advances coming from the research community, with agents trained to play Atari games from pixels or to be competitive with the best players in the world in complicated imperfect information games,” wrote the coauthors of a paper describing the work. “These systems have comparatively seen little use within the video game industry, and we believe lack of accessibility to be a major reason behind this. Indeed, really impressive results … are produced by large research groups with computational resources well beyond what is typically available within video game studios.”

The Ubisoft team, then, sought to devise a reinforcement learning approach that’d address common challenges in video game development. They note that data sample collection tends to be a lot slower generally, and that there exist time budget constraints over the runtime performance of agents.

Their solution is based on the Soft Actor-Critic architecture proposed early last year by researchers at the University of California, Berkeley, which is more sample-efficient than traditional reinforcement learning algorithms and which robustly learns to generalize to conditions that it hasn’t seen before. They extend it to a hybrid setting with both continuous and discrete actions, a situation often encountered in video games (e.g., when a player has the freedom to perform actions like moving and jumping, each of which are associated with parameters like target coordinates and direction).

The Ubisoft researchers evaluated their algorithm on three environments designed to benchmark reinforcement learning systems, including a simple platformer-like game and two soccer-based games. They claim that its performance fell slightly short of industry-leading techniques, which they attribute to an architectural quirk. But they say that in a separate test, they successfully used it to train a video game vehicle with two continuous actions (acceleration and steering) and one binary discrete action (hand brake), the objective being to follow a given path as quickly as possible in environments the agent didn’t encounter during training.

“We showed that Hybrid SAC can be successfully applied to train a car on a high-speed driving task in a commercial video game,” wrote the researchers, who futher noted that their approach can accommodate a wide range of potential ways for an agent to interact with a video game environment, such as when the agent has the same inputs as a player (whose controller might be equipped with an analog stick that provides continuous values and buttons that can be pressed to yield discrete actions through combinations). “[This demonstrates] the practical usefulness of such an algorithm for the video game industry.”

Let’s block ads! (Why?)

Big Data – VentureBeat

Non-Profit Uses Microsoft Azure to Advocate for Education Funding

December 14, 2019 CRM News and Info

The Commit Partnership Utilizes Microsoft Azure to Advocate for $ 6.5 Billion in Funding

Earlier this year, AKA Enterprise Solutions worked with The Commit Partnership, a Texas non-profit focused on education, to help them gain better insight into their data.

Since its inception, The Commit Partnership had managed to accumulate over 20 terabytes of data related to educational funding, student performance, and more. With so much data to account for, the organization was struggling with how to compile, analyze, and gain actionable insights from their data in order to make better and more strategic decisions.

Particularly when it comes to resource management and funding, The Commit Partnership relies on data insights to advise the schools they work with. The ability to report on prior successes is also heavily data dependent. Without this reporting in place, it’s difficult to secure critical funding from various partners.

At AKA Enterprise Solutions, we teamed up with Microsoft, DataKind, and Strive Together to develop a streamlined, easily accessible cloud platform for The Commit Partnership. By combining predictive analytics with the power of machine learning, the platform enables The Commit Partnership to run models against their vast data sets, offering new insight into their data.

The Commit Partnership Featured in a Microsoft Customer Story

AKA is thrilled that The Commit Partnership was recently featured in a Microsoft case study which highlights the non-profit’s use of Azure. Check out the case study here.

Is your organization looking to improve how you manage your data? Find out more about AKA’s non-profit focused solutions.

ABOUT AKA ENTERPRISE SOLUTIONS
AKA specializes in making it easier to do business, simplifying processes and reducing risks. With agility, expertise, and original industry solutions, we embrace projects other technology firms avoid—regardless of their complexity. As a true strategic partner, we help organizations slay the dragons that are keeping them from innovating their way to greatness. Call us at 212-502-3900!

Let’s block ads! (Why?)

CRM Software Blog | Dynamics 365

Nvidia uses federated learning to create medical imaging AI

October 14, 2019 Big Data

AI researchers from Nvidia and King’s College London have used federated learning to train a neural network for brain tumor segmentation, a milestone Nvidia claims to be a first for medical image analysis. The technique can allow data sharing between hospitals and researchers while preserving patient privacy.

Federated learning is an approach to machine learning that when using a client-server approach can eliminate the need to create a single data lake in order to train models. Rather, models are trained locally on devices then transfer insights from multiple machines to a central model.

“You need to get to these innovations, and I believe there’s kind of two ways. One, which we released last August, is create the best generalizable model that you have today, and just send it to each one of these hospitals where they can localize it for their own patients,” Nvidia director of health care Abdul Halabi told VentureBeat in a phone interview. “The other one is to say let’s fight together from the beginning, build this robust model as much as we can or generalizable model, and I think this research shows that it’s possible to actually do that here. It’s possible for you to achieve a high quality model without really bringing the data all together, which is why it’s really exciting.”

The model uses a data set from the BraTS (Multimodal Brain Tumor Segmentation) Challenge of 285 patients with brain tumors. The work will be presented at the Medical Image Computing and Computer Assisted Intervention Society (MICCAI) conference which begins today in Shenzhen, China.

“Most of the experiments that have been done is on synthetic data or just randomize data sets,” Halabi told VentureBeat. “But applying this was experimenting with real hospital data, using the brats challenge, to my knowledge there’s no work out there, that goes into the privacy direction.”

The potential impact of machine learning in health care is made evident today by the fact that some computer vision systems have demonstrated the ability to outperform human radiological experts, but diverse data sets with hundreds of thousands of cases aren’t always available due to heavy privacy requirements in health care.

That’s why, Nvidia senior research Nicola Reike said, many researchers in the field have used synthetic data sets or those compiled for challenges.

“So we’re saying that this research is really an important step towards the deployment of secure federated learning, and we hope that it will enable data-driven precision at very large scale,” she said.

The work explores the limits of differential privacy, a technique to add noise to data sets in order to make federated learning models more secure. Research has shown that without the application of differential privacy, neural networks can still leave some insights from underlying data exposed.

Apple and Google apply the same technique to federated learning for keyboard customization models on Android and iOS devices. Reike said federated learning for medical image analysis comes with its own set of challenges like the size of 3D medical images and need for much more compute power.

“We do this by injecting noise to each participating node, and this way it stores the updates and limits the granularity of the information that we actually share among the institutions,” Reike said.

Nvidia and King College’s work looks specifically at the ability to replicate models when portions are purposely hidden.

“If you only see let’s say 50% or 60% of the model updates, can we still combine the contributions in the way that the global model converges? And we found out Yes, we can. It’s actually quite impressive. So it’s even possible to aggregate the model in a way if you only share 10% of the model,” she said. “So it’s possible to only share 40% of the model. And we still reach the same accuracy or the same performance, as if the model were be trained on the pooled data.”

In other medical imaging news, the American College of Radiology (ACR) Data Science Institute said this spring it will incorporate Nvidia’s Clara AI toolkit into its platform.

Let’s block ads! (Why?)

Big Data – VentureBeat

Grammarly uses AI to detect the tone and tenor of your writing

September 24, 2019 Big Data

Nailing the right tone and tenor is of critical importance where writing’s concerned, whether the audiences of said writing are friends, coworkers, or hiring managers. The trouble is, without an extra pair of eyes, it’s rarely easy to know whether your work will have the intended effect.

Fortunately, Grammarly, a developer of cloud-hosted online grammar checking and plagiarism detection tools, has developed a tone detector the company claims can identify subtle contextual clues conveying a range of tempers. Essentially, it taps a battery of hard-coded rules and machine learning algorithms to spot signals in a piece contributing to its tone, including word choice, phrasing, punctuation, and even capitalization.

It’ll no doubt be of use to the 25% of respondents in a recent Grammarly survey who reported receiving feedback that their work email tone is too harsh. In that same study, 17% of young professionals under the age of 35 claimed they’d been told that their tone is too informal.

“Accurately identifying someone’s tone in writing is much harder than it is in conversation. For one thing, you have no facial expressions or body language to hint at a writer’s emotional state. For another, there’s usually no single attribute that is responsible for the tone of a message,” wrote Grammarly in a blog post. “In the future, you can expect Grammarly to not only help you identify the tone of your message but also offer specific suggestions for achieving the tone you want.”

To this end, tone detector, which is available in beta as part of Grammarly’s browser extension for Chrome ahead of a rollout on Safari and Firefox in the coming weeks, attempts to suss out the sentiment of any email, document, or blog post at least 120 characters in length. Recognized tones include “friendly,” “joyful,” “aggressive,” “annoyed,” and “excited,” among 35 others.

The new feature follows on the heels of a revamped interface that made it easier to keep tabs on consistent capitalization, spelling, acronyms, and more. Prior to that, Grammarly brought its AI-powered proofreading tools to Google Docs.

Grammarly — the brainchild of Kiev, Ukraine-based developers Alex Shevchenko, Max Lytvyn, and Dmytro Lider — was founded in 2009 shortly after the sale of Lytvyn and Shevchenko’s first startup, MyDropBox, to educational services company Blackboard. In May 2017, Grammarly raised $ 110 million from General Catalyst, IVP, and Spark Capital in its first venture round, and it has since grown to over 200 employees across offices in Kiev, San Francisco, and New York.

In addition to Google Docs, Office, and other platforms, Grammarly’s spell-checking tools are available on smartphones in Grammarly Keyboard, a keyboard replacement for iOS and Android that scans text messages, documents, and emails for errors in four English dialects: American, British, Canadian, and Australian.

Grammarly says its services are used by nearly 20 million people across all platforms. Its entry-level paid service, Grammarly Premium, costs $ 30 a month (or $ 140 a year), while its enterprise-oriented Grammarly Business tier is $ 15 per user per month. Both add style and vocabulary recommendations across all platforms.

Let’s block ads! (Why?)

Big Data – VentureBeat

How Skullcandy Uses Predictive and Sentiment Analysis to Understand Customers

September 19, 2019 Sisense

Mark Hopkins is the Chief Information Officer at Park City, Utah based Skullcandy, leading the global IT, Digital, and Customer Service teams. During Mark’s tenure, he has helmed Skullcandy’s digital transformation and his team has successfully increased online revenue and presence in the digital ecosystems, evolved into a world class customer service organization, and enabled growth with innovative systems solutions. Mark’s team is constantly adapting to and meeting the challenges of a rapidly evolving business using cloud technologies, real-time analytics, data warehousing, and virtualization.

Skullcandy’s journey with advanced analytics started with our product development team daring to ask three big questions:

What if we could predict return rates on new products before they were introduced?
What if we could use insights around reviews and warranty claims to understand positive drivers, negative drivers, and sentiment to inform new product design decisions?
What if we could use this data to focus our resources and deliver better products?

I often compare those early days to a kid standing at the edge of a bowl at the skate park — getting ready to drop in for the first time. He knows it isn’t going to be pretty, but he has to start somewhere. We knew our journey with predictive analytics and sentiment analysis was going to be a gradual progression that would eventually help us understand and better serve our customers. We knew we might end up with some bumps and bruises, but to gain the advantage we had to take that first leap of faith.

SQL Python R CTA 770X250 770x250 How Skullcandy Uses Predictive and Sentiment Analysis to Understand Customers

As a long-time Sisense customer, we knew that we had an open analytics platform on our hands that would be the backbone of this next analytics frontier for Skullcandy. Here’s how we dropped in to answer the questions our Skullcandy product team dared to ask.

Predicting Return Rates On New Products

The first piece of the puzzle was trying to see into the future — so we pulled BigSquid into the mix with Sisense to help with the heavy lifting. We were most interested in exploring if it was possible to predict the return rate on a new product based on historical return rates of products with similar features.

If you lack a data science team, integrating BigSquid with your open-platform BI tool is a powerful way to achieve the horsepower of data science while maintaining the ease of use that the average business user requires. We fed Kraken (BigSquid’s predictive analytics engine) information about historical warranty costs, claims, forecasts, historical product attributes, and attributes of the new products on the roadmap. Then we ran Kraken’s machine learning and predictive modeling engine to get the results.

Following a few false starts and some great iterative learning with the BigSquid team, we came away with a solid predictive data model of the warranty costs for future periods. That predictive output was then fed into Sisense so we could drill-down, explore, and use these predictions to make data-driven decisions, ask new questions, and understand our cost drivers. For our product development team, these kinds of insights are a goldmine for exploring opportunities for impacting warranty costs on new products before they’re even released.

Using Sentiment Analytics to Inform New Product Design Decisions

With our predictive data models telling us what might happen in the future with our products, our next step was to use sentiment analysis models to tell us what customers are saying and feeling right now. Again, with our BI housed within Sisense, we could integrate our text and sentiment data using a few different techniques. Our BigSquid partners suggested using Python and it’s natural language processing libraries to understand what customers are talking about, and Sisense helped us use AWS Comprehend to understand how our customers feel about our products.

With this integrated data tech stack, we could feed in text from customer reviews and warranty claims to be processed by a Python NLP Engine in order to pull out key themes. We could also feed the same data into Amazon Comprehend to measure the sentiment behind the themes, and which products were most associated with certain sentiments. Finally, we could load this robust sentiment analysis back into Sisense — turning what was previously a “needle in a haystack” exercise into a richly engaging data experience yielding very targeted analysis for our business users.

Where in the past we had to wade through disparate, siloed data, now we can correlate a stream of negative sentiment to reviews that mention a defect on the left side of a headset. And we could take it a step further to put a dollar value on that negative sentiment — by connecting those negative reviews on a product to the same product’s warranty claims. Incoming products with similar design and engineering could be given extra attention by our product designers and engineers before going to market. And we could easily visualize how a fix could impact our warranty claim forecast. Full circle data experience: achieved.

Lessons Learned

If you’re considering a similar investment in your company’s data strategy, there are a few lessons we learned along the way:

Be open to false starts with data modeling. Like any other data project, it won’t be instantly clear what your data model should look like. And bringing the predictive models into the mix means you’re connecting data points from the past, the present, and the “future” to build a model that provides actionable insights. It will be iterative. Be patient!
Make the data easy to add to and modify. For our data team at Skullcandy, that meant using an easy to manipulate view in SQL as our main data source. This allowed our team to easily swap out code and change the view as we continued to build and iterate.
You may not like what you see. It’s fun to explore the positive drivers and get lost in the positive reviews. One of the surprises that our NLP Engine identified in one product’s reviews was a predominance of the phrase “for my son.” It was fun to think about why that might be.
But we also came face-to-face with some harsh reviews. Not as fun to read — but when you invest in advanced analytics, sometimes that’s exactly what you came for. Your goal is to identify what isn’t working for your customers so that future products can deliver what they want and need.

Here at Skullcandy, we’re happy to report that “dropping in” to the predictive and sentiment analytics game was worth the initial uncertainty. We answered some of our most pressing questions, came up with some new insights we hadn’t originally considered, and this project has helped concretely demonstrate to the company what’s possible with advanced analytics. As a result, we are further along on our journey as a data-driven company.

Tags: predictive analytics | SQL

Let’s block ads! (Why?)

Blog – Sisense

AI startup Gather uses drones and computer vision for warehouse inventory

August 17, 2019 Big Data

Gather, a company that uses autonomous drones for warehouse inventory, launched out of stealth today. Founded in 2017, the company of about 10 employees is based in Pittsburgh.

Gather’s founding team is made up of Robotics Institute at Carnegie Mellon University graduates, including cofounder and chief robotics officer Sankalp Arora, whose work together with a team at the Office of Naval Research on autonomous helicopters won the 2018 Howard Hughes Award.

Earlier this year, Gather closed a $ 2.5 million funding round to bring its products to market and grow its computer vision and software offerings.

Gather supplies software for the autonomous operation of drones that can connect with existing warehouse management systems and IoT devices such as motion sensors. Computer vision is then used to scan and count inventory.

The solution is about 60% cheaper than traditional methods that rely on people alone, Gather said. Gather software is currently used in an undisclosed number of warehouses, but an air cargo company was able to reduce inventory time from 8 hours to 15 minutes, and take inventory 10x more quickly than two employees with a forklift in another warehouse.

“People never have to leave their desk to go out and do an inventory,” Arora told VentureBeat in a phone interview. “We’re not outfitting these things with any specific hardware. It’s a drone that you can go buy at Best Buy.”

Gather was founded in January 2017 by Arora, Daniel Maturana, Geetesh Dubey, and Robb Myer, an entrepreneur-in-residence at CMU.

Since Gather can connect with IoT devices, drones can deploy anytime a motion sensor is triggered. Arora argues a team of drones can multiply savings for warehouses by eliminating the need to install cameras for security or surveillance.

While drones using Gather’s computer vision can read barcodes and count boxes, an added bonus is providing hourly inventory image sets that warehouse managers can use to assess inventory reports or look back through records.

Cameras are also used to help the drones avoid running into objects.

“Our semantic mapping enables a single camera to map identified objects up to a range of 150 meters,” Arora said. “So what that means is we can follow rules of engagement in these facilities. For example, if a manager tells us ‘I don’t want this drone within 15 meters of a forklift,’ we can make a map these using semantic cues from known objects and map these objects by following rules of engagement.”

Computer vision by companies like Autodesk and Indus.ai are being applied in industrial environments to improve safety and compliance with regulations. Such systems may also play a role in the future to orchestrate the movement of people and machines as more robots enter warehouses. A number of challenges stand in the way of fully autonomous warehouses, Arora said.

“I think a good way to put it is you have to deal with unpredictable decision makers, as well as flexibility of the environment as you are making these things automated. And that is where I guess the final hill for robotics is,” he said. “You’ve got amazing robots making our cars, doing manufacturing on manufacturing lines, but as as soon as the environment becomes dynamic, it changes and you have to interact with humans or you have to be flexible given the environment, then the perception challenges associated with it. And the planning challenge associated with it becomes several orders of magnitude more difficult.”

Recent robotics ramp-ups include Brain Corp’s initiative to bring robots to warehouses and a $ 46 million funding round for Fetch Robotics in June, while DHL introduced a $ 300 million investment last year to quadruple robots in its facilities in North America.

Amazon, who rolled out its new Pegasus sorting robot in June, predicted earlier this year that fully autonomous Amazon warehouses are at least 10 years away.

The $ 2.5 million funding round was led by Expa and joined by Dundee VC, Bling Capital, Comeback Capital, XRC Labs, Summer League Ventures, and Plexo Capital.

Let’s block ads! (Why?)

Big Data – VentureBeat

Facebook open-sources image and video algorithms it uses to fight abuse

August 2, 2019 Big Data

Facebook today announced that it’s open-sourcing two algorithms capable of spotting identical and nearly identical photos and videos, which it says it actively uses to fight child exploitation, terrorist propaganda, and graphic violence on its platform. The company notes that it’s the first time it’s shared any media-matching technology — technology it hopes industry partners, smaller developers, and nonprofits will employ to more easily identify harmful content.

“When we identify a harmful piece of content … technology can help us find duplicates and prevent them from being shared,” wrote global head of safety Antigone Davis and VP of integrity Guy Rosen in a blog post time to coincide with Facebook’s fourth annual Child Safety Hackathon. “For those who already use their own or other content matching technology, these technologies are another layer of defense … making the systems that much more powerful.”

Facebook says that the two algorithms in question — PDQ and TMK+PDQ — were designed to operate at “high scale” and inspired by existing models and implementations, including pHash, Microsoft’s PhotoDNA, aHash, and dHash. The photo-matching PDQ was modeled after pHash (although it was designed from scratch), while the video-recognizing TMK+PDQF was developed jointly by the Facebook Artificial Intelligence Research team and academics from the University of Modena and Reggio Emilia in Italy.

Both efficiently store files as short digital hashes — unique identifiers — that help to determine whether two files are the same or similar, even without the original image or video. Facebook points out that these hashes can be easily shared among companies and nonprofits, as well as with industry partners, through the Global Internet Forum to Counter Terrorism (GIFCT), so they can also take down the same content if it’s uploaded to their services.

“We designed these technologies based on our experience detecting abuse across billions of posts on Facebook,” wrote Davis and Rosen. “We hope that by contributing back to the community we’ll enable more companies to keep their services safe and empower non-profits that work in the space.”

Facebook’s contributions of PDQ and TMK+PDQ follow on the heels of the aforementioned PhotoDNA 10 years ago, an effort to fight child exploitation. More recently, Google launched Content Safety API, an AI platform designed to identify online child sexual abuse material and reduce human reviewers’ exposure to the content.

Facebook CEO Mark Zuckerberg often asserts that AI will substantially cut down on the amount of abuse perpetrated by millions of ill-meaning Facebook users. A concrete example of this in production is a “nearest neighbor” algorithm that’s 8.5 times faster at spotting illicit photos than the previous version, which complements a system that learns a deep graph embedding of all the nodes in Facebook’s Graph — the collection of data, stories, ads, and photos on the network — to find abusive accounts and pages that might be related to each other.

In Facebook’s Community Standards Enforcement Report published in May, the company reported that AI and machine learning helped cut down on abusive posts in six of the nine content categories. Concretely, Facebook said it proactively detected 96.8% of the content it took action on before a human spotted it (compared with 96.2% in Q4 2018), and for hate speech, it said it now identifies 65% of the more than four million hate speech posts removed from Facebook each quarter, up from 24% just over a year ago and 59% in Q4 2018.

Those and other algorithmic improvements contributed to a decrease in the overall amount of illicit content viewed on Facebook, according to the company. It estimated in the report that for every 10,000 times people viewed content on its network, only 11 to 14 views contained adult nudity and sexual activity, while 25 contained violence. With respect to terrorism, child nudity, and sexual exploitation, those numbers were far lower — Facebook said that in Q1 2019, for every 10,000 times people viewed content on the social network, less than three views contained content that violated each of those policies.

Let’s block ads! (Why?)

Big Data – VentureBeat

ARCore’s Environment HDR mode uses AI to mimic real-world lighting

May 8, 2019 Big Data

The first day of Google’s I/O developer conference was predictably overstuffed with announcements about midrange Pixel smartphones, Google’s new camera-touting smart display, and Google Assistant’s forthcoming car-booking capabilities. Yet another item on the agenda was ARCore, Google’s software development kit for AR app development on Android and ChromeOS.

ARCore has been used to create thousands of apps to date and now supports an estimated 400 million devices (up from about 250 million devices in December), Google says. And this week, it’ll gain improvements on the rendering, design, and interactivity fronts.

First up: the Augmented Images API, which lets users point their cameras at 2D images (like posters or packaging) and bring them to life, can now simultaneously track moving images and multiple images. “This unlocks the ability to create dynamic and interactive experiences like animated playing cards where multiple images move at the same time,” wrote Google director of engineer Anuj Gosalia in a blog post.

The Light Estimation API, which provides a single ambient light intensity to mimic real-world lighting in digital scenes, as also been upgraded. A new mode — Environmental HDR — taps machine learning with a single camera frame to “understand high dynamic range illumination” in 360-degrees, taking in available light data and extending the light into a scene with accurate shadows, highlights, reflections and more. “When Environmental HDR is activated, digital objects are lit just like physical objects, so the two blend seamlessly, even when light sources are moving,” said Gosalia.

ARCore’s Environment HDR mode uses AI to mimic real world lighting

Above: Before and after Environmental HDR is applied to the digital mannequin on the left.

Image Credit: Google

Environmental HDR is available with three APIs:

Main Directional Light: helps with placing shadows in the right direction
Ambient Spherical Harmonics: helps model ambient illumination from all directions
HDR Cubemap: provides specular highlights and reflections

Last but not least, Google today debuted Scene Viewer, an AR experience that can be launched from the web instead of an app. It’s the framework that powers Google Search on mobile’s AR viewer.

Google says Scene Viewer and the other enhancements will roll out over the next two ARCore releases.

Let’s block ads! (Why?)

Big Data – VentureBeat

Tag Archives: Uses

Google’s Objectron uses AI to track 3D objects in 2D video

Ubisoft uses AI to teach a car to drive itself in a racing game

Non-Profit Uses Microsoft Azure to Advocate for Education Funding

The Commit Partnership Utilizes Microsoft Azure to Advocate for $ 6.5 Billion in Funding

The Commit Partnership Featured in a Microsoft Customer Story

Nvidia uses federated learning to create medical imaging AI

Grammarly uses AI to detect the tone and tenor of your writing

How Skullcandy Uses Predictive and Sentiment Analysis to Understand Customers

Predicting Return Rates On New Products

Using Sentiment Analytics to Inform New Product Design Decisions

Lessons Learned

AI startup Gather uses drones and computer vision for warehouse inventory

Facebook open-sources image and video algorithms it uses to fight abuse

ARCore’s Environment HDR mode uses AI to mimic real-world lighting

Recent Posts

Categories

Archives