Japan’s Alpaca launches Labellio, image recognition deep learning AI



See the original story in Japanese.

On June 30th Alpaca, a startup developer of AI (artificial intelligence)-powered apps and web services with office locations in San Mateo (Silicon Valley), Kobe, and Tokyo, launched a new platform called Labellio. The service can recognize and label images by leveraging their new deep learning technology.

Using Labellio pictures of gorillas can be differentiated from other images.

Logging into Labellio can be done using your GitHub or Google account. With Labellio users can take either their own images or search for images public on the web and sort them into multiple image groups with labelling. By repeating this process of labelling, the service learns from your actions and creates an image recognition model which can be either used in Labellio or can be exported as a data set for integration into other apps and products.

Alpaca expects that Labellio will be utilized in things such as automated tagging in photo sharing services like Google Photos, identifying someone by recognizing similar faces in multiple photos, recognition of weather conditions such as rain and snow using fixed-point live cameras, detection of a various brand’s products displayed in photos uploaded to social network sites, and other potential uses.

Alpaca co-founder and CEO Tsuyoshi Yokokawa explained:

The market we’re targeting is fairly a broad long tail. Social networks, community sites, image recognition and monitoring services, various areas. In the retail market, arrangements that work together with POS (point-of-sale) systems and recognize images of products are also a possibility.

We’ve introduced something similar to this before here on The Bridge, a service called Brand Pit, which can detect a brand’s logo in pictures posted on social media to gather information about their fans and customers. Unlike Brand Pit which is specialized for detecting brand logos, just by having Labellio learn through its deep learning engine it can draw data and perform labelling on what the product actually is, what kind of situation the product is in and other contextual elements of an image. It can also be conjectured that we may see in some industries buying and sharing of user-created vertical data sets that can produce image recognition intelligence, with a feeling similar to creating a specialized dictionary for translation systems.

Yokokawa added:

The types of things that Brand Pit is trying to do are also realizable with Labellio. Not just recognizing brands, with Labellio it’s possible to, for example, differentiate a picture of a cheeseburger from a picture of a hamburger.

We realized that there aren’t enough systems that people who want to do things like that can use, so we created Labellio so that it can be used by a lot of people.

According to Alpaca, they have developed Labellio as a sort of MVP (minimum viable product), and are currently developing another monetizable service based on the technology that Labellio uses, so we can look forward to the launch of that upcoming product.

Alpaca is the successor of Ikkyo Technology, a startup born out of Movida Japan’s 5th batch incubation program in 2014. Leveraging CEO Yokokawa’s work experience in New York and CTO Hitoshi Harada’s three and half years of experience as an engineer in Silicon Valley, Alpaca was established in the US to better serve the global market. They invited Luke Lonergan who is very active in the database community to join as an advisor, and are receiving support from organizations such as NEDO (New Energy and Industrial Technology Development Organization, Japan).

Alpaca also presented as a finalist in April’s Slush Asia pitch battle.

Translated by Connor Kirk
Edited by Masaru Ikeda