/ImageSearchEngine

Place recognition using Bag of Visual Words

Primary LanguageJupyter NotebookMIT LicenseMIT

Final Project of Computer Vision Course

Colab

Result

  • BoVW: 62.2%

  • ResNet-18 train from scratch: 76.19%

  • ResNet-18 Feature Extraction: 85.71%

  • ResNet-18 Finetune: 92.86%

  • NetVLAD (testing for place recognition)