Introduction and Preparation of the Dataset
In the first part of this tutorial series, you will learn what is classification and classification applications. Then, we will look at the first HDevelop example series on HALCON classification.
Within this program, we will learn how to read and split a dataset. Next, we preprocess the labeled data to be suitable for the deep learning model.
Then, you will be ready for training, which we will learn about in the next video.
0:00 Intro
0:22 Introduction of the technology
1:20 Working with the HDevelop example classify_pill_defects_deep_learning_1_preprocess.hdev
1:53 Read the dataset
2:33 Split the dataset for training, validation, and testing
3:30 Preprocess the dataset
Train a Model
In the second part of this tutorial series, you will learn how to train a classification model with MVTec HALCON.
We will have a look at some hyperparameters that influence the training progress, like for example the learning rate. Additionally, we will learn how and when to augment your data.
We will have a look at how to interpret the extensive training progress visualization. For example, we will examine the ‘top-1-error’, which is to check the performance of the model during training.
0:00 Intro
0:29 Select a suitable device for the training
0:53 Training hyperparameters: learning rate, batch size, epochs, …
2:47 Augment the training data
3:36 Visualization of the training progress
Evaluate the Trained Model – HALCON Deep Learning
In the third part of this tutorial series on HALCON’s deep-learning-based classification, we will check the deep-learning-based object detection model we trained in the previous video.
First, we will inspect some pie charts, visualizing the precision and recall of the model on a given dataset. Additionally, a confusion matrix helps us analyze the classes in detail. Finally, we have a look at the heatmap.
0:00 Intro
0:57 Evaluate the pie charts for precision and recall
2:48 Inspect the heatmap visualization
Apply the Model (Inference) – HALCON Deep Learning
In the last part of this tutorial series on HALCON deep-learning-based classification, we will apply the model we trained and evaluated before.
0:00 Intro
0:24 Select a suitable device for inference
0:48 Adapt the batch size for ideal performance
1:13 Inference workflow
About Us:
Advance Ultravision is the main distributor of Industrial Machine Vision products and components in Malaysia. We specialized in Machine Vision software such as Machine Learning, Deep Learning, Machine Vision Software Library, and 3D Machine Vision.
More information at advanceultravision.com
Curated By:
Vincent Hua is a Malaysia-based Machine Vision sales engineer, an ex-Machine Vision cum Automated Machine software engineer. Specialized in a machine vision system for Semiconductor / Electronics / Automotive sector.