Computer Vision based Transfer Learning from this project: https://github.com/kmock930/Texture-Image-Comparison.git
- Build 2 neural-network-based classifiers (i.e., ResNet50 and ResNet50-V2) to recognize more categories in the stonefiles dataset: https://web.engr.oregonstate.edu/~tgd/bugid/stonefly9/
- Perform Data Preprocessing.
- Evaluate the model's performance via some standard metrics and predictions.
- Perform Transfer Learning to learn a bounding box which accurately circles an object in an image.
- Perform Multi-Task Regularization and Data Augmentation.
- Evaluate the Transfer Learning process.
- Run the command
pip freeze > requirements.txtto generate the latest version ofrequirements.txtfile which keeps track of all necessary pip installs.
config.yamlis a configuration file that stores all parameters for modelling, training, experimenting, evaluating and etc.poc_hydra.pyis a standalone script to test the interaction with hydra in order to access (read/write) parameters in config.- Check out my Jupyter Notebook to see my analysis.