COMPUTER VISION & MACHINE LEARNING

A full description, use cases, and the necessary documentation can be found in the knowledge base.

Our models are built on the basis of OpenCV, TensorFlow, and other libraries.

OpenCV is an open-source library for computer vision, image processing, and general-purpose numerical algorithms.

TensorFlow is an open software library for machine learning developed by Google to solve the problems of building and training a neural network in order to automatically find and classify images, achieving the quality of human perception.

In your videos, CV determines both the objects and the probability of their detection. Each project has its own level of probability, ranging from a slight hint to the impossibility of the appearance of a specified object type. For example, a video has the tag EXPOSED_BREAST_F and a score of 0.51.

To determine the average value of your project, we recommend first taking a set of videos (for example, for a day or a week). Then, calculate the score of the specified tags for each video. Lastly, set coefficients based on the result analysis. For example, normal (max. 30%), questionable (max. 50%), and censored (51% and higher).

We operate with sets of images and videos that cover a large number of uses. However, the system sometimes needs additional training for particular cases.

We recommend generating a set of missed videos and sending them for analysis separately. In the next iteration, the system will also be trained on these videos.

Send images to the system for processing in the same way as videos. A picture is billed as a 1-second video.

The system takes into account the duration of each processed video in seconds. At the end of the month, the total number is sent to billing. The rate is calculated in minutes.

Let’s say you have uploaded three videos that last 10 seconds, 1 minute and 30 seconds, and 5 minutes and 10 seconds. The sum at the end of the month will be 10 s + 90 s + 310 s = 410 seconds = 6 minutes and 50 seconds. Billing charges 7 minutes. In your personal account, you can see a graph of minutes consumption for each day.

No. The streaming platform automatically deletes your videos and images after analysis and does not use your data to train basic models. Your video files do not leave your storage and are not sent to edge servers when the container is launched.