/Google-Landmark-Recognition

Image Classification and Image Retrieval on landmarks using transfer learning on VGG16 and DeLF. This is a famous Kaggle competition https://www.kaggle.com/c/landmark-recognition-challenge

Primary LanguageJupyter Notebook

Landmark Recognition

This code is for image classification on the famous Google Landmark Recognition challenge hosted on Kaggle.

Get started

  1. Download the data files 'train.csv' and 'test.csv' from here.
  2. Set up a GCP instance for optimal training time.
  3. Move on to the codes.

Have a look at the blog by our team to follow our journey.

Codes

The script 'Landmark_Recognition' consists of the code to

  1. Download resized images
  2. Move the images into folders labelled by class name
  3. Some visualizations
  4. Transfer Learning using Keras
  5. Predictions on test images

The script 'DeLF' consists of the code used for image retrieval done by matching local features.