What is Intra?
Intra is an application that can recognize image and behavior through machine learning. We have developed it for security purposes with video cameras. You can easily add cameras, create notifications when something suspicious is detected. The control room can view the interesting video streams on the overview page.
Learning on the job
Intra can learn just like a child because by giving an example which it can learn. Afterward, it can learn more and faster based on experience and better recognize behavior patterns or images. To demonstrate this, we created the video below with a still untrained Intra installation. We show you how it can learn a specific action and how well Intra is then already able to recognize behavior in other video material.
How does machine learning actually work with Intra?
There are two major forms of machine learning, supervised and unsupervised. With unsupervised learning, a lot of data is delivered to the computer and it recognizes patterns by seeing how often a certain combination occurs. For example, you can teach a computer a random language. By providing sufficient texts the computer will begin to recognize words and combination of words that belong together. Supervised learning works by labeling data. For example, take a set of photos where you have labeled the faces on the photos with the name of the person on them. The learning model must already be able to recognize objects such as faces or people and can then recognize them from the labels if a similar person is seen on a camera.
Recognition of persons of interest
Person recognition is an application where Intra is good to train for. This is relatively easy because you only have to provide a database with people (faces) and names. Extra information could be added to this, such as why someone is interested and what kind of notification should be sent when a person is detected. Learn more about Intra, or contact us for a demonstration on location.
How fast is machine learning with video material ?
Video contains is basically a long range of foto’s (frames) viewed in parallel. This makes video very good for unsupervised training purposes. We used this feature to train Intra faster. See how