Challenge

In a manufacturing context different parts must be assembled together. During this process there are some constraints on the orientation of movable parts. For example, the component shown in the images to the right is not allowed to have a deviation angle from the x-axis of the camera bigger than 15 degrees.

Solution

We trained a deep convolutional network for estimating different visual parameters including the deviation angle mentioned above, and other parameters like distance between predefined points or area of a particular surface.

Impact

The resulting solution allows for verification of defined project constraints and provides very accurate estimation results. For example, the test error for deviation angle is less than 0.6 degrees.