TransferLearning

In this project I predict 2 models. One where I predict whether an image contains a bird, squirrel or nothing. Another where I classify 102 flowers.

Purpose: The purpose of this project was to increase my practice in transfer learning & to use bayesian hyperparameter tuning for base-model & top-layer-mlp hyperparameters.

Results

Competition Results

• I placed 1st out of ~30 Machine Learning Engineers for the Birds v. Squirrels model.
• I placed in the top 10 for the flowers multi-class classification. ( I ran out of time however, and if I used similar techiques for BvS I would've achieved a much higher score on this model as well. )
• Too see confusion matrics and model results please see the /results directory.

Flowers Confusion Matrix

Birds v. Squirrels Confusion Matrix

Obtaining Data

• Flowers: The flowers dataset can be obtained through the script which calls the tensorflow datasets, specifically the: oxford_flowers dataset.
• BirdsVsSquirrels: The birds v squirrell dataset is a private dataset. If you wish to evaluate my models with the dataset feel free to reach out to me through this repository or through email.