Image search has got to be the obligatory burden that search engines need to carry around. In spite of the equally vast amount of spidering required, all that any self-respecting search engine can show are bare images strewn around, hopefully relevant. There seems to be absolutely no incentive for any service to throw up a decent set of image search results because in its native form, it is absolutely non-monetizable unlike its richer cousins like text or even other multimedia like audio and video.
Given that, the image search landscape was, by and large, filled with the big search players like Google and Yahoo! which indexed images by names and then threw out the best possible results. The primary incentive to serve relevant results was to siphon off most of these users to the more lucrative search categories and hence, gain a significant overall market share. Yahoo! went one up recently by exposing the vast database of images it possesses via Flickr as search results. Since the images are geo-tagged and presents readily relevant and indexed data, it was a natural progression to make them available as search results. Incentives as a result of this measure were added volumes in user engagement [capturing the user and his/her immediate social network via forwarding] and hence, a higher probability of the user conversions to signups for Flickr premium services.
The landscape, surprisingly, has seen mushrooming of a new set of players [like Polar Rose, Riya, Picitup] who base search results upon radically different approaches. A definite advantage these players enjoy is that these ventures are centered wholly around image based search and hence, their economics [or lack thereof] are not the same as those involving the traditional search engines. These services use the cheap availability of massive compute power to process images to determine patterns within them as against name/tag based approaches.
It goes without saying that relevance is of utmost importance; however, even bigger problems exist in terms of content and monetization. To counter that Polar Rose, being primarily centered around people based search, have introduced a Firefox plugin that helps bridge the content and relevance gap. The plugin scans the images loaded per page for human-colorized regions [much smarter of course] achieving near 100% accuracy in identifying faces in the image. That done, the plugin allows the user to tag the person as someone recognizable and the relevance is improved when that image is tagged again by someone else. Relevant results then get shown in the search results page. Overall, a healthy cycle that only allows better content and relevance as time goes by.
Riya has been around for some time and have recently introduced Like.com that aims to target shoppers who are visually attracted to certain products and on that basis, suggest similar products that might also appeal to the user. Tie-ups with small retailers and seeking a margin from the sales being driven by the visual search engine is, essentially, the revenue model being adopted by the venture. It remains to be seen how successful this turns out to be and how far it affects the profitability of the visual search space.
Picitup, meanwhile, is a late entrant in this field and employs a rather strange method of sourcing images, as explained in the Techcrunch article and which i have borrowed from below
An image search on Picitup begins with a textual search actually queried on Google or Yahoo. Picitup will display a set of results only from one of the two—the basis of the decision is the speed and quality of the results. The user can then select which image Picitup should fetch similar images for, or filter the results by Faces, Products, Landscapes and Color. The analysis is made in real time and is based on 100+ parameters including a propriety color space the company developed.
All said and done, its promising to see a number of players in the visual search space pitted against the traditional biggies and it would be interesting to see how the space pans out in the market share front. Picitup, as an initial arrangement towards that very goal, has introduced a celebrity match service that matches the facial impression of an uploaded image with those of celebrities and returns back those that match with an acceptable level of accuracy.
- Polar Rose as a facial recognition engine has been discontinued and has since been acquired by Apple.
- Riya.com and Like.com have discontinued their services and have since been acquired by Google.
- Picitup has since diversified to allow retailers bridge gap between online and offline shoppers by offering image based visual search as a platform to help find products.