• Facial Recognition

    Orbeus detects faces and facial attributes, such as race, emotion, age, and gender, and automatically groups the same faces to label friends and family.

    See Demo
  • Scene Recognition

    Orbeus determines the context and settings of images, automatically generates and tags searchable keywords.

    See Demo
  • Object Recognition

    Orbeus finds and tags animals, flowers and objects. With the labeled keywords, you can quickly find the images when you want them.

    See Demo


Orbeus labels your images with keywords, such as people, scenes, animals and objects, so that you can efficiently manage them and easily find them.


Orbeus leverages the Image-to-Text technology to index videos. We generate the metadata, as well as the associated timeframes, which can then be utilized to better match ads.

Our Products & Solutions


  • Image Data Mining: In the era of big data, people have taken so many photos everyday. Orbeus’ image-to-text capabilities are here to help you better analyze the photos and utilize the information.
  • Image Search: Orbeus’ contextual matching algorithm helps people to quickly search and find that perfect images, among hundreds of thousands of photos, simply by typing words.
  • Search by Keywords: Orbeus also provide you with the “search by image” capability. Sometimes, one image is equal to a thousand words. By using one image as search input, we will help you find more similar images.


Video Indexing

  • Movie Content Manager: highlight the appearance and time frames of celebrities, and interesting events in the movie.
  • Ads Match (Scene, Object, Animal, People Recognition): find related frames in the movie that can insert Ads.




How Image Recognition Works

When people look at a photo or watch a video, they immediately identify what’s in the image—people, animals, objects, brands and sceneries. This level of recognition capabilities were nearly impossible for computers, until now. Orbeus’ revolutionary image recognition technology helps computers to see like human beings.

Provide An Image

Given an image, we first scan the image and perform an universal object, logo, face and text detection.


Then we abstract features of the objects from the image. In this step, we find the candidates of objects in this image.

The Neural Network

The neural network is trained to be able to recognize a set of objects. After the neural network is trained,when it looks at the object candidates, it will generate the contextual content, meaning it will tell what the object is, e.g., face, girl, car, tree, ocean, street, city, nightlife, etc.

Image to Text

Based on the outputs of the neural networks, the semantic tags can be generated to comprehensively describe the content of the image.

Attribute Analysis

After we identify the objects from the first part, the second part is deep dive and attribute analysis.

Model Construction

The core part here is the model construction. If we want to recognize President Obama, we need to construct a model for Obama. The model construction is based on features abstracted from real images. That’s what the 3d facet is about: the features of images.

Model Comparison

After the model is constructed, we can compare an input image with a model.

Connect With Us

If you’d like to receive the occasional update or announcement from Orbeus, please sign up for our newsletter:

Get In Touch

Thank you for visiting Orbeus! If you’d like to reach out to us, please send us a quick message using the form below: