Machine learning algorithm to locate o-ring component in engineering drawing using small dataset. O-ring component are modelised to increase dataset. Aim of this study is to evaluate the success of 3 different models and to assess the score for each.

Segmentation algorithm with CNN (Convolutional neural network).

Project runs on Google Colab.

Dataset is created from parts of drawing with size 500x500px by adding modelling o-ring. With this, we can switch from 563 training data to more than 7,500.

Example of input and output of model. Algorithm find and locate correctly 2 o-rings.

• mylib.ipynb contains all common functions used in oring_projet.ipynb
• oring_projet.ipynb is the file with main function.
• oring_optim.ipynb is the file to tune hyperparameters