- Forevery not only classifies objects into a total of 11K possible categories – such a finite classification is a feat in itself – but also detects abstract concepts such as ideas, emotions, and feelings. E.g. you can find photos that portray the subjects being relaxed.
- Like the traditional face recognition features from the old desktop age in programs such as Picasa, Forevery enables the user to classify objects with a specific tag name. E.g. if a rocking chair was originally grandma’s, you can tag it as such and the system will identify in real time “grandma’s rocking chair” in all your photos. I.e. in Forevery user-based identification of clustered images is not limited to human faces.
By Hans Hartman – Suite 48 Analytics
On Monday I reported on the demise of Dropbox Carousel, pointing at two possible culprits (syncing is a poor substitute for backup; the new battleground is not photo aggregation/syncing but image recognition).
To illustrate the latter, today two image recognition vendors made exciting announcements: Clarifai launched its Forevery image recognition app and Ditto added Objects and Context to its Ditto Brand Manager and Social Stream API. [update: and if these two announcements are not enough, Pinterest also announced this morning that they acquired Pext, an app that helps you find photos to match your text messages and then easily turns them into memes]
Image recognition API vendor Clarifai released its first consumer app, offering several innovative features:
- Not all searches have to be conducted by typing in text that corresponds to how the images are tagged: Forevery also lets you perform visual searches by submitting a photo, after which you find photos that portray similar content.
- Clarifai’s deep learning roots shine in how Forevery learns from how you share particular photos with particular friends. It uses these learnings to make suggestions at a later time for which photos to share with whom. Think Google Now for photo sharing.
Ditto Objects and Context
Having a database of 6500 logos as a proprietary asset for classifying brands in in user-generated content is one, but finding images with specific logos in specific settings is another.
Today Ditto added Objects and Context functionality to its API, allowing brand owners to find social media photos of their branded products with particular form factors (e.g. you can search for a specific logo on a drinking glass, plastic bottle, or glass bottle), or in particular settings (in a refrigerator, vending machine or at a super market).
In other words, brand identification of UCG imagery is becoming more finite.
With Google announcing its Cloud Vision App (more about that in a future post) and the types of innovations like those from Clarifai and Ditto described here, as well as the batch tagging features from SnapsBoard announced a few days ago, the potential of solutions that bring photos at your fingertips is still largely untapped.
I look forward to seeing more exciting innovations coming to market in the next few months!