Adobe today announced a slew of updates and integrations across Adobe Document Cloud, its suite of cloud-based apps for converting and processing PDFs. The company’s free document scanning app for Android and iOS, Adobe Scan, is getting some improvements thanks to machine learning. Adobe Sign, its electronic signature service, is now deeply integrated with Microsoft Dynamics and offers access to LinkedIn customer details, plus a self-service tool for fielding GDPR requests. And starting this week, enterprise and team Acrobat DC users who subscribe to Office 365 can create PDFs from the ribbon menu in the web-based versions of Microsoft Word, Excel, PowerPoint, OneDrive, and SharePoint.
“Whether onboarding an employee, signing up a new customer, or completing a critical sales contract, great experiences start where the document does, in Adobe Document Cloud,” Ashley Still, vice president and general manager at Adobe, said in a statement. “As leaders in document and productivity software, Adobe and Microsoft are integrating best-in-class cloud services like Microsoft Office 365, Microsoft Dynamics, Adobe Sign, and now Adobe Acrobat DC to meet the needs of today’s agile and rapidly evolving workforce.”
Ahead of today’s announcement, Adobe gave VentureBeat an in-depth look at the enhancements coming to Adobe Scan. The app, which was developed by Adobe’s San Jose and Noida, India offices, has been downloaded more than 10 million times since it launched in March 2017. It leverages the power of Sensei, Adobe’s in-house machine learning platform, to convert documents, pictures, business cards, and whiteboards to PDF documents.
Before today’s update, Scan could automatically detect borders, sharpen content, and apply optical character recognition (OCR) to make text editable. The new version adds a one-touch button that saves business card info to a contact list.
“There’s still a big challenge with paper to digital,” Lisa Croft, group product marketing manager at Adobe Document Cloud, told VentureBeat in a phone interview. “Paper still remains in over 80 percent of digital processes. We’re helping enable customers to act on [those] processes whenever the opportunity arises.”
Once Adobe Scan recognizes a business card, it presents a Save Contact button. Tapping on it adds the person’s name, company, phone number, email address, and image to your contacts, and if you’re an Adobe Document Cloud subscriber, it will save a copy in your cloud storage locker.
But Akhil Chugh, product manager at Adobe, said the performance of its system is unparalleled.
Adobe Scan taps the heterogeneous compute capabilities of both Android and iOS devices to remove unwanted objects from the scan, like fingers, and uses heuristics models to identify fields like email addresses with 99 percent accuracy. In the rare instances it doesn’t get a name or number right, it offers machine learning-powered suggestions.
Since the app’s launch almost a year ago, the team has been continuously retraining its AI models with new data using TensorFlow, Google’s machine learning framework, improving their accuracy.
“There’s a whole bunch of levels of different complexities when you’re scanning with a [smartphone] camera,” Chugh said in a phone interview, “but [executives] are always on the move from one meeting to another. They don’t want to carry business cards around — they want to quickly archive them and digitize them.”
From a computer vision perspective, business cards fall roughly into three categories, Chugh explained: business cards with black text on a white background, which are the most common; black cards with white or gray text; and colored cards with color backgrounds.
It didn’t take long to train Adobe Scan’s models on the first category, but the second — black business cards — initially threw them for a loop. Improving scan accuracy required inverting the color of black business cards after the fact, in software (to the AI models’ eyes, the inverted cards appear to have white backgrounds with black text).
Equally challenging was getting the system to recognize white business cards on white tables and tablecloths. “It took a couple of months to solve,” Chugh said.
Low-light scans were the other beneficiary of the team’s data crunching. In environments where the ambient lighting is too dim for the camera to make out the business card’s text, the Adobe Scan app now briefly switches on the flash at the moment of capture. Shadow detection has been improved, too — Adobe Scan’s algorithms can distinguish between business cards and hard shadows formed by wayward hands and phones.
The new Adobe Scan is available for Android and iOS devices starting today.
Office 365 integrations
In September 2017, Microsoft partnered with Adobe to make Adobe Sign the “preferred e-signature solution” for its customers. Expanding on that collaboration, Sign and Adobe PDF are gaining new integrations with Microsoft Dynamics and Office 365.
Specifically, Microsoft Dynamics customers can now pull customer data from LinkedIn Sales Navigator and embed Adobe Sign into Dynamics workflows. In addition, Adobe Sign has been granted authorization from the Federal Risk and Authorization Management Program (FedRAMP), an assessment and authorization program to which U.S. federal agencies must adhere, and includes a privacy administrator role that includes the aforementioned GDPR request tool.
On the Office 365 side of things, new shortcuts in the ribbon of Word, Excel, and PowerPoint web apps allow you to convert documents into PDFs with optional password protection. Adobe PDF archiving, conversion, and distribution tools are also now available from within OneDrive and SharePoint.
The Adobe PDF integrations are available today with Adobe Acrobat DC for teams and enterprise, and the Adobe Sign for Dynamics 365 is available with Adobe Sign for Enterprise.
Adobe says that more than 200 billion PDFs were opened in Adobe Products in 2017, and that 50 percent of Fortune 100 companies now use Adobe Sign.