During the past year Autodesk's Forge Team has worked on new services made available to the public with the mission to improve the set of SASS products that will power the Future of Making Things.
A few months ago, we introduced Reality Capture API, which provides the photogrammetry capability to process digital images into high resolution textured meshes, dense point clouds and orthophotos. This highly scalable and easy-to-use API uses the power of the cloud to compute structure-from-motion and multi-view-geometry algorithms. Any REST-capable mobile, desktop or web application can connect to this API. One of the biggest industry applications has been the use of Drones in the Construction industry, which has become a game changer technology.
Just recently we released Webhooks API. the idea of webhooks is that sends data to endpoints (URLs) when a certain event occurs. It is triggered by events occurring in web applications. It then sends real-time data to applications listening to it. Since the data is sent immediately, using a webhook is more efficient that frequently polling for updates.
Similarly, the Forge Webhooks API allows applications to listen to Forge Data Management events, and receive notifications when they occur. When an event is triggered, the Forge Webhooks API sends a notification to a callback URL you have defined. You can customize the types of events and resources to receive notifications for. For example, you can set up a webhook to send notifications when files are modified or deleted in a specified hub or project
Today, I’m excited to announce the new set of services of Forge Large buiLding Image RecOgnitiOn API’S (Hopefully you guys knew this was a joke - April Fools), This new API can easily detect a broad sets of objects in your images, from thousands of Buildings to other object categories commonly found in not Buildings. These API’s are a new concept that introduces advance image accuracy and learns overtime to recognize your beautiful buildings.
This BETA API is still under development and only a few early adopters have been given access to start developing the next generation of Image recognition apps powered by Forge.
Here is a sample test of how the services work. A user opens his mobile camera to snap a picture of an object, in this case a Building in NYC. After the API is done with the analysis a label will be displayed showing what the object in the picture is.
The JSON response of the API call looks like this.
"labelAnnotations": [
{
"mid": "/m/0cgh4",
"description": "building",
"score": 0.96018714,
"topicality": 0.96018714
},
{
"mid": "/m/05_5t0l",
"description": "landmark",
"score": 0.92486346,
"topicality": 0.92486346
},
{
"mid": "/m/079cl",
"description": "skyscraper",
"score": 0.8856122,
"topicality": 0.8856122
},
{
"mid": "/m/03nfmq",
"description": "architecture",
"score": 0.8122836,
"topicality": 0.8122836
},
{
"mid": "/m/056mk",
"description": "metropolis",
"score": 0.79981047,
"topicality": 0.79981047
},
{
"mid": "/m/020ys5",
"description": "condominium",
"score": 0.7902488,
"topicality": 0.7902488
},
{
"mid": "/m/0vlys",
"description": "tower block",
"score": 0.6930415,
"topicality": 0.6930415
},
{
"mid": "/m/01fdzj",
"description": "tower",
"score": 0.6330253,
"topicality": 0.6330253
},
{
"mid": "/m/03c8phm",
"description": "corporate headquarters",
"score": 0.61261195,
"topicality": 0.61261195
},
{
"mid": "/m/01x314",
"description": "facade",
"score": 0.6068641,
"topicality": 0.6068641
},
{
"mid": "/m/01bqvp",
"description": "sky",
"score": 0.60465974,
"topicality": 0.60465974
},
{
"mid": "/m/026y56",
"description": "brutalist architecture",
"score": 0.5629683,
"topicality": 0.5629683
},
{
"mid": "/m/05hg54",
"description": "commercial building",
"score": 0.5553216,
"topicality": 0.5553216
}
]With this data we are able to create a better UI/UX for our customers to understand better what the image being analyzed is.
Let’s try with a different object. How about an old school way of new york city transportation.
For this case we do not have very accurate data at the time of the analysis yet, but we are working on making this better.
Currently the BETA API does a recognition between 2 types of objects, but we know is the beginning of a long journey ahead of us and we couldn’t be more excited about it. Also another current problem are reflective objects.
We know Frank Gehry would be very upset but a fixed is being worked on and a push to PROD will soon be released.
The project has been added to our Github repository as MIT License, grab a copy and contribute to this Open Source project. Here is the direct link to it.
Forge Large buiLding Image Recognition API
Hope you guys enjoy what we are creating here, and look forward to hear your feedback about it.
Happy Easter! and Happy April Fools!