XImage Trainer
XImageTrainer is free software for preparing image data and collaborating with other deep learning frameworks to train datasets. Without a good GUI tool, drawing bounding boxes for objects in thousands of images would be a painful job. XImageTrainer provides basic tools to make this task more convenient and simple.
Download the latest release
Notice: Windows 10 users may see the Microsoft Defender SmartScreen warning dialog the first time they open it. You can ignore this issue temporarily by pressing [More info] -> [Run anyway].
Currently supported deep learning frameworks
1. Setting the Darknet
Go to Darknet to obtain the source code, then build Darknet, which uses the YOLO network to train image datasets.
Click the [Setting] button to open the Settings dialog.
Choose Darknet's executable file and command arguments.
2. Create a new project
Click the [New Project] button to open the Add Project dialog.
Enter the project name and workspace directory. After generating, the project structure consists of Classes, Configuration, Train Images, and Test Images.
3. Add class (category) name
Right-click on [Classes] and choose [Edit Class Categories] to open the Edit Class dialog. Enter a new class name for each row.
4. Configure YOLO network
Right-click on [Configuration] and choose [Edit Configuration] to open the Configuration dialog.
5. Add Images and ROI (Region of interest)
Right-click on [Train Images] or [Test Images] and choose [Add Image] to open the Open File dialog. You can select one or multiple files to import. Images are copied into the project's Images folder. Images that already exist cannot be added again.
Click on Image to open it in the center image view. Choose [Rectangle] on the toolbox and draw bounding box for object.
After drawing the bounding box, a Class Selection dialog will pop-up. You need to choose the right class for this object and click [OK] to confirm.
Notice: all the images must be drawn bounding box before starting the training.
6. Run the project
Click [Run] button to Build & Start training the project. After finish without any error, you can find the final model file in the project's Backup folder.
