Run main.py with the following arguments to query an image using the base CBIR system. All parameters are optional. By
default, the first image in the University of Washington dataset (./data) is set as the query image.
The best 10 matches are retrieved by the system, displayed and written to an automatically created (./DIRNAME_out)
folder. The folder is emptied for each result-set.
python3 -m pip install --user virtualenv
python3 -m venv venv
source ./venv/bin/activate
pip3 install -Ur requirements.txt
python3 main.py --query_dirname query --query test.jpg
Run main.py with the following arguments to change the number of returned matches, or the second-stage comparison
distance measure threshold.
python3 main.py --query_dirname query --query test.jpg --matches 20 --threshold 20000
Run main.py with the --delay flag to increase the viewing time of the query image from the default three seconds.
Note that the delay is specified in milliseconds.
python3 main.py --query_dirname query --query test.jpg --delay 5000
Run main.py with the --pca flag to use PCA to reduce the dimensionality of the feature vectors. The feature vector
database, if it does not exist, will be saved to disk as DBNAME_pca.npz.
python3 main.py --query_dirname query --query test.jpg --pca
Run main.py with the --kdtree flag to store the
python3 main.py --query_dirname query --query test.jpg --kdtree
Run main.py with the --horizontal flag to adjust the weight w_{12} by 20%. This elevates the horizontal image detail
in the comparison procedure. The default adjustment factor is 1.5, or 50%
python3 main.py --query_dirname query --query test.jpg --horizontal --scale 1.2
Including the --vertical, --diagonal or --intensity flag will similarly adjust the corresponding weight. All flags
and combinations thereof can be specified.
Run main.py with the --help flag to view the Help file.
python3 main.py --help
usage: main.py [-h] [--dbname DBNAME] [--dirname DIRNAME] [--query_dirname QUERY_DIRNAME]
[--query QUERY] [--matches MATCHES] [--delay DELAY] [--threshold THRESHOLD] [--vertical]
[--horizontal] [--diagonal] [--intensity] [--scale SCALE] [--pca] [--kdtree]
Content Based Image Retrieval with Wavelet features
optional arguments:
-h, --help show this help message and exit
--dbname DBNAME Database name
--dirname DIRNAME Image database directory name
--query_dirname QUERY_DIRNAME
Query image directory name
--query QUERY Query image filename
--matches MATCHES Number of matches to return
--delay DELAY OpenCV window millisecond delay
--threshold THRESHOLD
Euclidean distance threshold
--vertical Emphasise vertical image detail
--horizontal Emphasise horizontal image detail
--diagonal Emphasise diagonal image detail
--intensity Emphasise intensity variation
--scale SCALE Weight scaling factor
--pca Perform feature dimensionality reduction
--kdtree Store features in K-d tree
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent