Application of computer vision technology to seismic facies classification
Application of computer vision technology to seismic facies classification
This project involves applying computer vision technology to seismic interpretation
The code for this project is available here
The aim of this project is to improve the acccuracy of Reinaldo et al (2019). Please feel free to contribute
SSD Mobilenet is the choice architecture for this project. The following methods will be attempted to improve the accuracy of the model
- Using exponential moving average (EMA) from tensorflow_addons
- Label smoothing
- Using Stochastic Gradient Descent with warm restarts.
- Tweaking regularization parameter
- Data augmentation for very similar images
- trying the MobileNetV3
- Jittering