HP Labs has launched an interesting set of APIs for facial recognition, image matching, and image extraction. Named the “HP Labs Multimedia Analytic Platform”, the APIs leverage HP’s experience in digital image processing (http://map.hpl.hp.com/). Our team came across the platform while searching for a facial recognition API. Having heard about HP’s focus on cloud computing, we decided to take one of these APIs for a spin and find out how easy it would be to build an app on this cloud platform over a weekend.
On the HP Labs site we found an Image Collage API that sounded interesting. It takes a selection of photos, identifies key vectors (i.e. faces) in the photos, and creates a collage of the photos, keeping the key vectors visible in the collage. We decided to combine this API with the Facebook API to let users automatically generate a collage from a selection of their Facebook photos. The collage can be set to various shapes. Since Valentines Day wasn’t far off, we chose a heart shape to help our friends who wait until the last minute to come up with a gift.
We started building the app Friday at 6 pm. The HP Labs documentation was succinct and the code examples clear. What we found especially helpful was HP’s explanation of the technology powering the collage creator, and the integrated Sandbox. These shortened the time required to get up to speed on using the API, and were helpful when considering what UI functionality to build in.
On Saturday we built out the UI and integrated the Facebook API. That evening we began testing the HP Cloud performance with live runs using our photos from Facebook. We had minimal expectations given that this HP Labs API is still in Beta. It was a pleasant surprise to find out that the API performed well (since the initial app launch. Uptime for the API has not been 100% but HP has responded to our support requests when needed). We used HP cloud object storage to hold the uploaded photos which integrated seamlessly with the collage API.
On Sunday we completed the UI, named the app “KissAPI” (www.kissapi.com), and completed the app website. We conducted several more systems tests to verify performance and with all systems go, we launched the app after one weekend of work. Since launch, the HP support has been satisfactory considering the Multimedia Analytic Platform is a beta program within the company. Overall it was a good experience with HP’s Cloud and we plan to use the API again to build additional apps.
About the authors:
Garland Kan and Andy Bartley are on the founding team of Algorithms.io. Algorithms.io is building a cloud service to help data scientists and developers discover and use new algorithms. Learn more at www.algorithms.io.
- Technology & Electronics