casting-defect-computer-vision

Casting Defect Detection using Computer Vision

This application demonstrates the usage of the Kelvin SDK to implement a solution to detect manufacturing defects using computer vision and machine learning.

The solution consists of two main components:

1. Camera Connector: Simulates image capture from a camera and publishes the image in base64 format to the Kelvin Platform. In production environments, it would interface directly with live camera feeds to acquire real-time images.

2. Casting Defect Detection: Processes the acquired images to identify casting defects using a pre-trained TensorFlow machine learning model. It evaluates the images for any anomalies and reports the findings back to the Kelvin Platform for further analysis.

Architecture Diagram

The following diagram illustrates the architecture of the solution:

Additional Resources

• Dataset: The dataset employed for training the model is hosted on Kaggle and can be accessed here.

• Pre-trained Model: The Tensorflow machine learning model used for defect detection is also available on Kaggle. It can be found here.

Requirements

1. Python 3.8 or higher
2. Install Kelvin SDK: pip3 install kelvin-sdk
3. Install project dependencies: pip3 install -r requirements.txt
4. Docker (optional) for upload the application to a Kelvin Instance.

Usage

1. Upload both applications to a Kelvin Instance: kelvin app upload
2. Deploy the Camera Connector application
3. Deploy the Casting Defect Detection application