This is a CBIR system implemented by Rails and Python
Video demo: https://youtu.be/mNyZ0TLwdP0
- Ruby version: 2.5.1
- Configuration image data
Please put your image in the database folder like this structure
- Prepare cache data for the further search
python3 src/resnet.py
(This command will generate the cache data in folder cache and data.csv in the root directory. If you update your image dataset and want to update the cache data, you need to remove cache folder and data.csv)
- How to run the query test
python3 query.py image-you-want-to-search image-category
Example:
python3 query.py public/1558249242.jpg effiel
Search Result
Using cache..., config=resnet152-avg, distance=d1, depth=3
{'img': 'public/1558249242.jpg', 'cls': 'effiel', 'hist': array([1.1803220e-03, 1.2102447e-03, 1.8271050e-04, ..., 4.8219634e-05,
1.0576390e-03, 5.9930864e-04], dtype=float32)}
database/eiffel/paris_eiffel_000000.jpg 0.1455633044242859, Class eiffel
database/general/paris_general_000084.jpg 0.3631804883480072,Class general
database/general/paris_general_000138.jpg 0.3734038174152374,Class general
database/general/paris_general_000143.jpg 0.3752489686012268,Class general
database/general/paris_general_000021.jpg 0.3753284215927124,Class general
database/eiffel/paris_eiffel_000005.jpg 0.377275288105011, Class eiffel
database/eiffel/paris_eiffel_000217.jpg 0.37977999448776245, Class eiffel
database/general/paris_general_000144.jpg 0.38824301958084106,Class general
database/eiffel/paris_eiffel_000136.jpg 0.4028390049934387, Class eiffel
database/eiffel/paris_eiffel_000128.jpg 0.40601611137390137, Class eiffel- Reference
https://github.com/ry/tensorflow-resnet
https://github.com/KaimingHe/deep-residual-networks
https://github.com/pochih/CBIR/blob/master/src/resnet.py
http://web.stanford.edu/class/cs276/handouts/EvaluationNew-handout-1-per.pdf