Pimloc origins
THE BEGINNING
2012: The genesis of Pimloc began with the creation of autonomous (‘intelligent’), wearable, image capture devices; within a previous business, OMG Plc. These devices automatically captured a natural first-person documentary of hobbies or social interactions, allowing a user to focus ‘on the moment’.
The devices used a range of sensors to determine when to capture an image or video, based on changes within the environment: such as changes in a users movement, light levels, ambient temperature, scene content and location. Users loved being unshackled from living behind their camera lenses, but had to trade that against sifting through streams of noisy content after the event, to find the moments they cared about.
2013: Our first device, Autographer, was launched - closely followed by Google Glass and a host of other wearable imaging devices, which raised questions around the widespread capture and usage of personal data. These discussions continue even now; with the growing adoption of smart home devices equipped with cameras and microphones, and the large-scale roll-out of CCTV and other surveillance technologies.
These devices surfaced a range of new data privacy considerations for ‘always-on’ visual capture. Whilst responsible capture, storage and usage of content was designed into the Autographer product by default:
all content was stored locally on device and PC/laptop
sensitive meta-data was only viewable through the desktop application
the camera lens had a visible privacy screen so that others could easily see when it was switched off, and
devices came with an etiquette guide for responsible usage
The process of taking this product to market highlighted a wider need for a more holistic approach to the way personal data would need to be captured, stored and managed in the future. Data protection legislation has since developed with a range of data protection laws, including GDPR, which now provides a framework for the management of personal data. Legislation that is now spreading from the EU into the US and beyond.
2014: In parallel with data protection challenges, was how to deal with the so called ‘wall of death’ (endless streams of noisy images and video). Being free in the moment was offset against spending lots of time after the event sifting through streams of content to curate the best highlights.
This reviewing and editing challenge inspired Pimloc’s resulting visual classification technology. The journey started with personal connections being made with the principal researchers at Oxford University’s AI lab (Visual Geometry Group); Professor Andrew Zisserman FRS and Professor Andrea Vedaldi. They were already investigating how to create deep learning systems for the automatic ingestion and classification of large scale visual data in order to extract wider scene context.
Our early developments demonstrated the power of deep learnings systems for visual classification tasks; alongside providing early thinking into wider privacy and security responsibilities for dealing with such systems and data.
2016: As a result, Pimloc was formed (spun out of OMG Plc), in order to develop a visual classification system that would allow individuals and businesses to harness the power of deep learning for images and video whilst protecting personal data.
Since then: Pimloc’s codebase and IP has developed through a range of projects, market specific engagements and close working with academia. The resulting solutions have come together to create an underlying technology platform for the mass ingestion and classification of visual content which is delivered through two main products:
1. Pholio: a search and discovery system for large scale visual content
2. SecureRedact: a SaaS video privacy platform for selectively anonymising personal data