Cutting Inserts Inspection

Machine vision algorithm utilized for damage inspection of the cutting inserts used in lathe machine. Project utilizes hybrid approach to the image classification by using "clasical" and deep learning algorithms. Program is prepared for smart camera ADLINK NEON 2000 more_info .

Table of Contents

• General Info
• Technologies Used
• Features
• Project Development
• Setup
• Room for Improvement
• Google Colab

General Information

• Project enables vision controll of the cutting inserts usage
• Main pourpose is to provide autonomy of the controlling process
• Hybrid approach - "classical" and deep learning methods

Technologies Used

• Python 3.6.9
• TensorFlow 2.3.1
• OpenCv 4.3.0
• NumPy 1.18.5
• SciPy 1.6.2
• JetPack 4.4 with following software enviornment more info here .

Features

List the ready features here:

• Autonomously finding cutting edge
• Detecting breaches of the cutting edge
• Distracting surface contaminations from real damages

Project Development

Setup

Project utilizes cutting inserts images collected by ADLINK NEON 2000 camera. Stored: https://drive.google.com/drive/folders/1lf42JO9PQdO09VsaYtjwALDH3pwM55xS

Room for Improvement

• Algorithms optimazation.
• Providing greater independence from lighting.
• Examining other deep learning pre-trained networks
• Feeding deep learning algorithm with syntetic data generated in Blender

Google Colab

Used for preparing neural network model by transfer learning and utilizing Incepction V3.