With the MVTec Deep Learning Tool it’s possible to train a deep-learning-based classification model from scratch.
First, we import images of your application – using the folder structure, the Deep Learning Tool can assign labels automatically. We can then easily review our data by filtering the images according to these labels. After that, we train the model using transfer learning and pretrained networks provided by MVTec. Lastly, we can review the trained model in detail, using a confusion matrix, heatmaps and more. In this video, version 0.5 of the Deep Learning Tool is used.
The exported trained model including the associated dictionary, which contains the preprocessing parameters, can be imported into HDevelop and MERLIC and then used for inference.
In this video, version 0.5 of the Deep Learning Tool is used.