Upgrade to Pro — share decks privately, control downloads, hide ads and more …

[CVPR20 Tutorial] Live-coding Demo to Implement...

[CVPR20 Tutorial] Live-coding Demo to Implement an Image Search Engine from Scratch

[CVPR20 Tutotrial] Image Retrieval in the Wild
https://matsui528.github.io/cvpr2020_tutorial_retrieval/

Live-coding Demo to Implement an Image Search Engine from Scratch
Yusuke Matsui

video: https://www.youtube.com/watch?v=M0Y9_vBmYXU
code: https://github.com/matsui528/sis
demo: http://www.simple-image-search.xyz/

Yusuke Matsui

June 16, 2020
Tweet

More Decks by Yusuke Matsui

Other Decks in Research

Transcript

  1. 1 Live-coding Demo to Implement an Image Search Engine from

    Scratch CVPR 2020 Tutorial on Image Retrieval in the Wild Yusuke Matsui The University of Tokyo
  2. 2 Implement an Image Search Engine from Scratch ➢ Search

    by deep-feature ➢ Web-interface ➢ Pure Python Demo: http://www.simple-image-search.xyz/
  3. 3 ➢ A variety of open-source libraries are available, making

    it easy to implement and deploy an image-processing system ➢ Demo to implement a content-based image search engine ➢ Only 24 lines (html) + 43 lines (python), w/o comments Implement an Image Search Engine from Scratch Code: https://github.com/matsui528/sis Demo on AWS EC2: http://www.simple-image-search.xyz/ +
  4. 4 offline online 0.21 0.43 0.02 0.59 offline.py Feature Extractor

    (Keras, VGG16) 0.17 0.84 0.22 0.86 … server.py Feature Extractor Web Server (Flask) 0.42 0.73 0.31 0.07 0.21 0.43 0.02 0.59 0.17 0.84 0.22 0.86 … Browser Search . ├── feature_extractor.py ├── offline.py ├── server.py ├── static │ ├── feature │ │ ├── img01.npy │ │ └── img02.npy │ ├── img │ │ ├── img01.jpg │ │ └── img02.jpg │ └── uploaded │ └── query01.jpg └── templates └── index.html img feature