This is an automatic detection, segmentation, and counting of mechanical parts app based on Mask RCNN and Kivy.
If you want to reimplement this to your own task, please read tutorials for Mask_RCNN. For example Balloon example or Shapes example.
Common steps for making the decision:
- Prepare data for your task.
- Install necessary packages (Keras, Kivy, Numpy, Matplotlib, Opencv etc.)
- Use VIA annotator to annotate objects on images.
- Run Jupyter notebook, change paths to images, begin training of the neural network and validate results.
- Implement weights of your model (.h5 file) into the graphical interface. Just lay down .h5 file in root directory of this app.
- If you want to build your app for Windows or Linux, use .spec file and pyinstaller package.
Raw Image
After NN processing
Interface example
