Industry Casting Defect Classification

Global Casting Industry of Metals s around 200 Billion USD. But the main problem in this industry is Casting Defects, which occur due to main parameters like temperature, shape, material, etc. Casting defects result in increased unit cost and lower morale of shop floor personnel. Unfortunately, this is not an easy task, since casting process involves complex interactions among various parameters and operations. Types of Casting Defects are are

• Filling related defect
• Shape related defect
• Thermal defect
• Defect by appearance

Our project will definitely helpful to increase the productivity and yield of the castings.

Introduction:

We are using a CNN to train a Dataset of Defective and Non-Defective Images of Metal Casted Product. And to Infer the Trained Model we are using OpenVino to increase speed of testing

Colab Notebook Link

Benefits:

• It will automate the Quality Analysis in Manufacturing Industry
• It will speed up the process and decrease errors
• It will increase Production Rate and Decrease Cost of Defect Correction
• It will analyse the Defects in Casted Products and help industrialist

Social Impact:

• Save Time
• Decrease Labour Cost
• Increase efficiency and Energy Output

Use of OpenVino:

I have used OpenVino Toolkit 2020 to infer the dataset from trained model.

Training

The Model was Trained for 25 Epochs and getting Accuracy of 99%

Inference on Normal CPU :

The Inferences were taken on images of Test Data from Dataset

Results :

Time Taken Per Image : 0.288 Seconds

Inference using OpenVino Toolkit:

All Codes are written and Tested on Google Colab for Easy Testing

Installation:

Windows - link

Linux- link

OpenVino was used on Final Trained Model

Steps:

1. Converting.h5 model to.pb (frozen model)

2. Converting.pb (frozen model) to.xml and.bin file required for OpenVino.

3. Running the final inference model using OpenVino's Python Libraries.

The Inferences were taken on 50 images of Test Data from Dataset

Results :

Time Taken Per Image : 0.102 Seconds

Conclusion:

As you can see OpenVino Clearly gives boost to Inference by 250%, I have used OpenVino for my previous projects and it is very easy to use and very much efficient.

Our project will definitely helpful to increase the productivity and yield of the castings. It can be easily deployed in industry Vision System by using RaspBerry Pi and Intel Movidius Stick. Currently I didn't have those hardware due to lockdown so I didn't Made the Original Hardware for implementation.