Image recognition has come a long way, even since we published our report “How apps can solve photo management” just a year ago. At this point major imaging, storage, and social media vendors like Dropbox, Yahoo, Facebook, Google, Pinterest, and Shutterfly have all acquired image-recognition startups, and they are pursuing the holy grail of understanding what is shown in a consumer’s photo or video so that this imagery can be automatically categorized, retrieved, equipped with content-sensitive links, or otherwise leveraged.
Today’s consumers need the ability to locate relevant photos in their ever-expanding collections. Respondents in our survey assessed solutions for these needs to be valuable but mostly unavailable in the marketplace. While some image-recognition solutions cater to these consumer needs, others focus on the needs of advertisers and ecommerce vendors, who benefit from providing suggestions and links that are aware of image content, similar to how they have also leveraged the analysis of social media texts for advertising and sales purposes. We believe the consumer-driven needs, coupled with the resources and motivation of the advertising-focused social media companies, will provide the image-recognition cross-market breakthroughs that requirements in asset management and stock photos, retail, health care, manufacturing and robotics, and security — or academia — have failed to deliver.
Those consumer offerings will build on academic deep-learning technology, which uses neural networks and massive computing power to create and refine image-recognition algorithms. Image recognition is not yet at the level of voice recognition or OCR and its accuracy varies widely, but given the fast progress, we expect it to get there for most use cases in the next 12 to 36 months. Non-consumer sectors should monitor and adopt the technologies driven by these innovations.
Further progress around image recognition will come from:
• Developing training sets for more classification categories
• Leveraging additional data sources
• Optimizing photos prior to image recognition
• Expanding beyond identifying objects or people
We believe that image recognition will have a considerable impact on many markets, in particular:
- Consumer photo organizing. Image recognition gives new photo-organizing services the opportunity to replace traditional photo-organizing tools.
- Photo services. Image recognition benefits the photo-output industry, as it enables consumers, who are increasingly overwhelmed by the sheer number of their dispersed photos, to still find the pictures worth printing. It also provides opportunities for newcomers in the micro or UGC stock-photo markets to compete with incumbents who still rely on traditional photo-organizing methods.
- Web and mobile advertising and ecommerce. With consumers paying attention to UGC photos, image recognition enables advertisers and ecommerce providers to place content-sensitive links on or near photos.
- Enterprise asset management. Image recognition provides enterprises the tools to track the use of their branded assets in social media so that they can determine the effectiveness of their campaigns or any misuse of their visual assets.
Feature image courtesy Flickr user Lubomir Panak
- Consumer requirements and advertising drive innovation
- Recognition technology poised for disruption
- Applying image recognition
- Market impact: ripple effect from consumer services
- Key takeaways
- About Hans Hartman
- About Gigaom Research