InstanceSearch

Instance search assignment for CS4186 Computer Vision & Image Process

Experiment reproduction

Best method: SIFT

• Customized file path and parameters (explained in report) in utils.params.py
  • DATASET_PATH = project working directory (root directory)
  • QUERY_DIR = query image directory
  • DATA_DIR = dataset directory
  • GT_DIR = ground truth directory
  • DES_DIR = directory for descriptor files
  • COL_DIR = directory for color files
  • RET_DIR = SIFT result directory
  • LBP_RET_DIR = LBP result directory
  • LBP_DIR = directory LBP feature files
• Generate descriptor files for image i(i = 0, 1, ..., 5019)
  • Interface: from utils.fileIO import generate_descriptor_file
• Retrieve top-10 Match
  • Interface: from SIFT.Interface.match import match_main
• Ranklist
  • Interface: rank_list()

Last but not least, if you have problems in generating descriptor files (do not know how / extremely long running time), please feel free to reach out to me for descriptor files generated in my experiment. Unfortunately, I cannot upload zip file of them on canvas due to huge file size...

Email: [email protected]