GoogleLandmarkRetrieval

This is my solution for Google Landmark Retrieval Challenge

Description is available at here

Requirement

• python
• keras
• cv2
• faiss
• download image data from here and place them in input/index/ and input/query
• place sample_submission.csv in input/

How to use

• run extract_feature_vgg.py
  extract feature at dir output/
  you can change a layer to extract features by change line 57
• run output_submission.py
  calcurate distances between features and output submission file at dir output/
  you can change a layer to extract features by editing line 7
  you can change a model to extract features by editing line 8
• run train_rotnet.py
  train RotNet
• run extract_feature_rotnet.py
  extract feature at dir output/
  you can change a layer to extract features by change line 78