Image Classifier Bronze

Picnic Hackthon 2019

Intro

The model is build with Keras and uses InceptionResNetV2 as the base model. The base model is fine-tuned and trained by using k-fold on the training set.

Steps to reproduce a .tsv file:

• The Download folder from PicNic 'The Picnic Hackathon 2019' should be unzipped in this folder.
• Add all test images inside './data/test/no_labels'
• Run the first 6 cells inside 'picnic_train_model.ipynb'. This creates all the necessary folder structure.
• Run the 'picnic_predict_model.ipynb' code. The actual model is inside the 'models' folders.

Steps to create a new model:

• The Download folder from PicNic 'The Picnic Hackathon 2019' should be unzipped in this folder.
• Run the 'picnic_train_model.ipynb' model.

Requirements:

• Python 3.5
• Jupyter Notebook

Python requirements:

• numpy
• pandas
• matplotlib
• keras
• PIL

Built With

• jupyter-notebook
• keras
• python