Linux gizmo indexes photos and videos for visual recognition search
Nov 20, 2017 — by Eric Brown 738 viewsPimloc’s “Pholio” runs Linux on an Nvidia Tegra, and provides offline storage and search of images and video using visual and face recognition.
Digital imaging has lived up to its promise of making it easier to take more images more quickly, but the promise that it would make it easier to find those images has fallen short. Unless you spend time with an image management package and apply tags to each and every photo, it’s a pain to try to find specific images or groups of images. A new Kickstarter project called Pholio promises to skip the prep work and use visual recognition technology to quickly locate any image or video you seek.
![](/files/pimloc_pholio-sm.jpg)
![](/files/pimloc_pholio_ui-sm.jpg)
Pholio (left) and mobile app showing search UI
(click images to enlarge)
The tabletop, 235 x 195 x 60mm Pholio device connects via WiFi, Ethernet, or USB to your phone, computer, storage system, or online storage site, and collects all your photos and videos. It then indexes thousands of images, and provides visual and recognition technology, including face recognition, to search by keyword or image match. The Linux-based technology offers mobile and desktop apps, but is not linked to any cloud service — it operates completely offline, enabling user privacy.
The Pholio is available through Dec. 9 on Kickstarter for 299 Pounds ($396) with 500GB storage or 399 Pounds ($528) for a Pro model with 2TB, due in January and February, respectively. It’s an all or nothing project, and so far the project has achieved only 20 percent of the $126,250 goal. Yet, even if the campaign fails, we imagine we’ll see something similar show up from Pholio in the future.
— ADVERTISEMENT —
The relatively high price — unless you were one of the lucky 50 who got the 199 Pound early bird — reflects software development as much as the hardware itself. The Pholio is offered by an eponymous London-based startup spun off from visual deep learning firm Pimloc, which collaborated with University of Oxford scientists on the device’s AI algorithms. The software was built on a training process that sorted through millions of images, says Pholio.
The Pholio lets you choose between making a full resolution copy of an image or video, or saving space by storing an optimized version. In the latter case, images are compressed to the 3MB to 4MB range, and video shrinks to 720p resolution.
Using the optimized thumbnails, you can store 140,000 images on the 500GB model and 560,000 on the 2TB Pro model. You can even convert RAW files to the compressed format while maintaining a link back to the original.
![](/files/pimloc_pholio_scuba-sm.jpg)
![](/files/pimloc_pholio_multi-sm.jpg)
Pholio subject search (left) and multi-person search feature
(click images to enlarge)
Once images are indexed, you can search using 20,000 search terms, or click on an image to bring up all images related to that person. A new feature mentioned in the KS campaign’s Update tab, enables searches based on multiple faces in an image so you can bring up images that include all the faces you select.
Neural network algorithms are available for face, object, and scene recognition and classification, and the algorithms can be trained to reflect the whims and patterns of users. For example, you can swipe right or left for “more like that” or “less like that” control, and users can create their own “specialist classifiers.” Pholio can search for relevant frames or segments within videos. You can also link the Pholio with your Internet browser to search for images.
![](/files/pimloc_pholio_randall-sm.jpg)
Pholio CEO Simon Randall and a possible ancestor identified by Pholio
(click image to enlarge)
The technology has been tested on several large, specialist collections in British museums, galleries, and libraries. Simon Randall, CEO of Pholio, tested the device on an extensive fine arts image library at the Courtauld Institute at Somerset House. He asked Pholio to search for a lookalike of his own image within the library “and in a fraction of a second retrieved this image of a possible ancestor,” says Pholio. (We imagine celebrity lookalike searches will be a major Pholio pasttime.) In addition, Pholio claims that a search for a lizard within the movie “The Terminator” quickly displayed “a split-second scene early in the film with a pet lizard is placed on the top of some shelves.”
![](/files/pimloc_pholio_detail-sm.jpg)
Pholio detail view
(click image to enlarge)
Pholio has not supplied many hardware details. However, in an email reply to LinuxGizmos, Pholio reps said the device runs Linux with OpenEmbedded on an Nvidia Tegra SoC (type unstated). Other features include a microSD slot, as well as Ethernet, HDMI, and USB 3.0 ports. The WiFi uses WPS security, and more extensive cryptographic security is “coming soon.”
![](/files/capsulelabs_capsule.jpg)
Capsule
If you’re looking for a more affordable offline image backup solution that also offers some image search features, you may want to check out the Capsule Kickstarter campaign from Australia based Capsule Labs. The device is available in early bird packages starting at only AU$ 119 ($90), with deliveries in May 2018. The spherical device has no onboard storage, however, and according to a lengthy Pholio reply to a KS commenter, the Capsule image recognition software is not as sophisticated as that of the Pholio. The Capsule campaign is halfway to its $49,017 goal, and will run through Dec. 10.
Further information
The Pholio is available through Dec. 9 on Kickstarter for 299 Pounds ($396) with 500GB storage or 399 Pounds ($528) for a Pro model with 2TB, due in January and February, respectively. More information may be found on the Pholio Kickstarter page and Pimloc’s Pholio website.
Please comment here...