App | deepProfile

our goal

There are numerous variants of aluminum, steel, plastic and rubber profiles on the market, which may differ only slightly in their geometric shape and size. Our app deepProfile is based on machine learning algorithms and allows different types of profiles to be captured and immediately catalogued uniquely in a server and thus identified.

1: Picture

You can now easily identify the profiles in your warehouse or in production using a camera on your smartphone. You can also select previously captured images from the gallery and identify them accordingly through the app.

2: Analysis

The analysis of the profiles uses a cloud-based neural network. The combination of Skel Net and Deep Hash technologies lead to an exceptional accuracy rate.

3: results

The results of the analysis are displayed directly with the corresponding matching rate. It is also possible to visualize multiple results with the respective matching rate.

reliable

The app is particularly reliable because it is based on a combination of two technologies. Currently, more than 6000 profiles are stored on the server.

quick

In less than 3 seconds, the app shows up to 5 match results with the corresponding hit rate.

easy-to-use

The app is designed to be very user-friendly, in just a few steps you will get the result.

adaptive

Whenever the server is asked to identify a profile it does not know yet, it replaces the missing knowledge and continues to grow to become more and more sufficient.

Compatibility

All major smartphones and browsers are supported (Android, iOs, Windows, ...).

precise

The app can correctly analyze even the most complicated photos.

Technology

deepProfile is based on the processing of image data using a neural network based on Deep Hash and Skel Net technologies. By using this combination, deepProfile can
provide highly accurate and fast results and minimize the risk of errors.

Deep Hash

Deep Hash is used in image processing and pattern recognition. It is capable of reducing high-dimensional data to a compact form that can be easily compared.

This technology generates a representation of images that allows similar images to be quickly identified. For this purpose, a CNN (Convolutional Neural Network) is used, which allows to create an abstract representation of the image. This representation is then converted into a sequence of bits called a hash.

Unlike other image processing techniques, Deep Hash aims to convert similar images into a similar hash. This allows similar images to be quickly identified by simply searching for similar hashes. In the hashing approach, each image is represented as a vector of features and then a binary code is generated from it. The great advantage of hashing is that the classes can be stored, retrieved and compared in the database very efficiently.

Skel net

Skel Net was developed specifically for the analysis of 2D and 3D shapes. It is able to recognize and classify complex shapes and patterns.

This technology uses a method called "skeletonization" or "topology extraction" to convert an image into a simplified model that represents the basic shapes of the object. This method allows complex shapes and patterns to be recognized and classified.

The neural network is trained by a large amount of training data to identify patterns and features in the shapes. Once the network is trained, it can be used to analyze new images and recognize the shapes and patterns in those images. Deep learning using the skeleton principle is widely established for detecting and representing objects. By means of the thickness of the skeleton, information about the contour can also be stored efficiently.

"We are proud to bring this new technology to market," says a spokesman for e. Luterbach AG. "It will help to simplify and optimize work processes in various industries. We are convinced that deepProfile will provide great added value for our customers."