Watson Visual Recognition in 2018
Get started with Watson Visual Recognition through IBM Cloud.
From JetBlue using facial recognition to replace boarding passes to H&M embracing image search, image recognition saw widening appeal in 2017 as newcomers and nontraditional customers embraced computer vision capabilities. This was fueled in-part by the staggering trends in photography: over 1.2 trillion photos taken worldwide in 2017, 85% of which were captured on smartphones.
As the Watson Visual Recognition team continues to introduce new capabilities and tools that make it easier for everyone to transform their businesses with computer vision, we want to thank all of our users for their support, many of whom have been with us since we first launched in May 2016. From inspecting rusty towers using drones to recognizing shipping containers to preventing consumption of Tide Pods, our users have shown endless creativity in applying Visual Recognition towards challenges of all shapes and sizes.
Below is a recap our service achievements from this past year and as well as a sneak peek into our plans for 2018.
2017 Recap
Color Model (Beta)
In early 2017, we introduced the Color model as a new beta feature. While Watson had already been able to detect color, the Color model returns the top colors it saw in each image as response tags, each accompanied by a classification score. The capability allows users to quickly assess the dominant colors within images and turn these into actionable insights.
Tooling (Beta)
In April 2017, we introduced a new developer tool that enabled users to create and manage their Custom models through a web app. By entering a Visual Recognition API key, users could use our GUI to seamlessly create, retrain, and delete Custom models associated with their API key without needing to go through the hassle of forming complex HTTP requests. Since all actions were tied to the user’s key, any edits made via command line would be reflected in the tooling and vice versa.
Food Model (Beta)
In May 2017, we introduced the Food model as another new pre-trained model. While Watson had already been able to detect food as part of the General model, the Food model provides enhanced specificity and accuracy for food items based on a vocabulary of over 2,000 tags.
Explicit Model (Beta)
Introduced in December 2017, the Explicit model classifies whether an image contains nudity or sexual content that would be inappropriate for general use.
General Enhancements
Throughout the year, a number of enhancements were made to the service more broadly. Image size limits were increased to 10 MB and language support was expanded to 7 languages. The General model also saw a number of improvements, including greater scene detection capabilities, enhanced object recognition, and an increase in the average number of returned tags per image to 10.
Looking Ahead: 2018
The Watson Visual Recognition team has a number of exciting plans in place for 2018. While we can’t announce everything just yet, our goals include improvements to both service features (including highly-requested capabilities not currently part of the service) and user experience. The latter category includes tools that will make it easier for users to utilize and implement Visual Recognition in a number of ways, as well as behind-the-scenes work that will improve processes such as support ticket handling.
We’ll be making a number of announcements at Think 2018, IBM’s flagship annual conference happening in Las Vegas from March 19–22. Register for the conference here or learn more about how to snag a VIP ticket here.
If you’ll be at the event, feel free to send me a note and come say hi at one of our sessions:
- [8065] Automating Inspections with Belron
- [8073] AI for Good (ft. Stadtreinigung Hamburg)
- [3197] Digital Forensics with IBM Chief Analytics Office
- [8617] Hands-on Lab: Inspections
- [8670] Hands-on lab: Visual Recognition + iOS
Demo booth on 2nd Floor — DevZone
Questions? Comments? Working on your own Watson projects? Let us know in the comments below, and don’t hesitate to reach out!
