Perfect Half Million Beauty Product Image Recognition Challenge

This is my solution for Perfect Half Million Beauty Product Image Recognition Challenge , which obtained the 2nd place (GAN) with MAP@7 0.29135.

Requirement

• python 2.7
• pytorch 0.4.0
• PIL
• torchvision here
• pretrainedmodels here
• faiss here
• download image data from here if published and place them in ./data/ , ./val/ and ./test/
• place val.csv and test.csv in ./

How to use

• run python extract_all_feature.py
  extract feature at dir ./feature/
  • -imsize (default:480) : the size of longer image side used for extracting feature
• run python retrieval.py
  calcurate the MAP@7 between the data features and the val or test feature, the submission file at dir ./
  • -iseval (default: 1) : whether eval with the MAP@7
  • -impath (default: './data/') : the file for data
  • -tpath (default: './val/') : the file for val data, if the data is test data, change to './test/'
  • -tlabel (default: './val.csv') : the label for val data, if the data is test data, please set -iseval as 0