Introduction to MVTec Deep Learning Tool

Introduction to MVTec Deep Learning Tool

The MVTec Deep Learning Tool V0.4.3 has added model training and evaluation features based on its image classification and object detection annotation capabilities. This feature helps customers train deep learning models and provides relevant evaluation information, including heatmaps of predicted categories and confusion matrices for inference images. In the DeepLearningTool (referred to as DLT), users can click on the evaluated confusion matrix to observe the classification status of each category. Meanwhile, the new version of DLT also provides inference time statistics, allowing users to better understand the inference performance of the model.

Introduction to MVTec Deep Learning Tool

DLT Model Training Function

After double-clicking to open the DLT tool, users can create a specified project in the new project tab, taking the classification project as an example:

Introduction to MVTec Deep Learning Tool
Introduction to MVTec Deep Learning Tool

After completing the creation of the classification project, select the add image folder button (using the HALCON-provided pill image folder as an example).

Introduction to MVTec Deep Learning Tool

Select the root directory of the images and click the “Add Folder” button.

Introduction to MVTec Deep Learning Tool

After adding the image folder, users can see that the DLT tool has categorized the images according to the folder levels.

Introduction to MVTec Deep Learning Tool

In this tab, click the “Import and Annotate Images” button to import the sample dataset. After doing so, you will see that all six tabs: “Project”, “Gallery”, “Images”, “Review”, “TRAINING”, and “EVALUATION” are now activated.

Introduction to MVTec Deep Learning Tool

Project: Preview previously opened DLT projects

Gallery: Preview all images in the current project

Images: Preview the currently selected image

Review: Preview annotation information for the entire dataset

TRAINING: Provides training functionality for the current project

EVALUATION: Provides model evaluation functionality

Clicking on the training tab will show the following GUI information:

Introduction to MVTec Deep Learning Tool

Among them, “DATASPLIT” indicates the sample distribution status, and customers can divide the dataset according to actual conditions. Here, samples are divided in a ratio of 70, 15, 15.

After the sample division is complete, hyperparameters for this training process can be configured in the “SETTING” tab:

Introduction to MVTec Deep Learning Tool

It is worth mentioning that DLT V0.4.3 supports training on both CPU and GPU platforms, and the pre-trained models have added Alexnet and mobilenetv2 on top of compact, enhanced, and resnet-50. Users can choose the appropriate model for training based on their needs.

Introduction to MVTec Deep Learning Tool

Switching to the RESULTS tab allows you to see the training process:

Introduction to MVTec Deep Learning Tool

Clicking the left arrow symbol allows you to export the trained model:

Introduction to MVTec Deep Learning Tool

EVALUATION – Evaluation

After training, you can evaluate the test set in the EVALUATION tab:

Introduction to MVTec Deep Learning Tool

In the CONFUSION MATRIX, you can see the output confusion matrix. It is noteworthy that the confusion matrix provided by the DLT tool is interactive, allowing users to click on the contents of the confusion matrix to observe the evaluation status of the model.

Introduction to MVTec Deep Learning Tool

If you need to purchase or try the product, please call the Daheng Imaging customer service hotline: 400 999 7595. We will continue to provide you with reliable products, advanced technology, and attentive service!

Contact Information

Hotline: 400-999-7595

Daheng Imaging (Group) Co., Ltd. Beijing Imaging Visual Technology Branch Website: www.daheng-imaging.com Email: [email protected] Headquarters Address: 12th Floor, North Building, Daheng Technology Mansion, No. 3 Suzhou Street, Haidian District, Beijing (100080) Switchboard: 010-82828878

Shenzhen Office Address: Room 1601, Desai Technology Building, No. 1 Gaoxin South 1st Road, Nanshan District, Shenzhen (518040) Phone: 0755-83479565Shanghai Office Address: Room 1808, Hongxing World Trade Building, No. 598 Nujiang North Road, Putuo District, Shanghai (200333) Phone: 021-35312826Wuhan Office Address: Room 2009-2010, Block A, Guanggu Times Square, No. 111 Guanshan Avenue, Hongshan District, Wuhan (430079) Phone: 027-87223690Xi’an Office Address: Room 1601, Block F, Wangzuo Modern City, No. 35 Tangyan Road, High-tech Zone, Xi’an (710065) Phone: 029-89521122Chengdu Office Address: Room A508, China Resources Plaza, No. 5 Chengyi North Street, East Second Ring Road, Jinjiang District, Chengdu (610021) Phone: 028-86925034

Xiamen Office

Address: Room 803-1, Building B03, Software Park Phase III, Jimei District, Xiamen

Phone: 0592-5500803

Northeast Region Agent Harbin Sanhe Micro-nano Technology Co., Ltd. Address: Room 1617-1618, Teaching Electronics World, West Dazhi Street, Harbin (150001) Phone: 0451-86262530

Introduction to MVTec Deep Learning Tool

← Official WeChat Account

↓ Official Douyin Account

Introduction to MVTec Deep Learning Tool

Leave a Comment