Manual

Table Of Contents

DobotVisionStudio User Guide

149

Deep Learning is a machine learning algorithm, which is developed based on the traditional neural

network. With a deep learning module, the machine can absorb, learn, understand and handle

complex information in the real world as human beings.

Specifically speaking, machine with the deep leaning module, can execute and finish high-difficulty

recognition tasks, such as character recognition, character positioning, image segmentation, image

classification, object detection, image retrieval, anomaly detection and instance segmentation.

As deep learning is based on data, you need to collect a large amount of data sets before starting

deep learning. The data sets participating in the training need to be labeled to ensure the diversity

of data as much as possible and have low resolution requirements. Take character positioning as an

example. At least 150 picture samples are required. You need to use VisionTrain for deep learning

training.

NOTE

 The deep learning module has two versions: G and C. The suffix of module name of

GPU version is G, the suffix of module name of CPU version is C, and the CPU version

does not depend on the graphics card.

 DL character recognition, DL code reading, DL character location, and DL single

character recognition belong to the recognition category.

 The function module has a high requirement for the PC configuration. See Operating

Environment for details.

DL Image Segmentation

In image segmentation, you can configure parameters and detect foreground objects through self-

trained model. Image segmentation can be divided into two-level classification mode and multi-

level classification mode. The two classification mode is to only detect defects, as shown in the

figure below.

• Two-level Classification Mode