DIY image retrieval with spatial verification
Project description
Hands-on wide baseline tutorial
Summary description here.
We will create the wide baseline stereo mather and try it on various images with various features. There is also a a (naive) example of the spatial verification together with image retrieval. We will not build the components from scratch, instead will be using a ready packages, like kornia, pydegensac and OpenCV
Install
pip install local_feature_tutorial
How to use
%matplotlib inline
import numpy as np
import matplotlib.pyplot as plt
from local_feature_tutorial.wbs import *
from local_feature_tutorial.visualization import *
from local_feature_tutorial.io import *
from local_feature_tutorial.datasets import *
import cv2
hard_images_to_match = 'http://cmp.felk.cvut.cz/~mishkdmy/wbs_illum.tgz'
fname = download_file(hard_images_to_match)
untar_to(fname, 'data/wbs')
wbs_img_fnames = get_all_images_in_subdirs('data/wbs')
print (wbs_img_fnames)
visualize_grid(wbs_img_fnames, (8,8))
['data/wbs/chimera_01.png', 'data/wbs/chimera_02.png', 'data/wbs/dh_01.png', 'data/wbs/dh_02.png', 'data/wbs/doll_theater1.jpeg', 'data/wbs/doll_theater2.jpeg', 'data/wbs/doll_theater3.jpeg', 'data/wbs/kn_church-2.jpg', 'data/wbs/kn_church-8.jpg', 'data/wbs/ministry_01.png', 'data/wbs/ministry_02.png', 'data/wbs/ministry_03.png', 'data/wbs/purkine-2.jpg', 'data/wbs/purkine-4.jpg']
sift_hardnet_matcher = TwoViewMatcher(detector=cv2.SIFT_create(8000), descriptor=HardNetDesc(),
matcher=SNNMMatcher(0.9), geom_verif=degensac_Verifier(0.5))
res = sift_hardnet_matcher.verify(wbs_img_fnames[7], wbs_img_fnames[8])
print (res.keys())
draw_matches(res['match_kpts1'], res['match_kpts2'],
wbs_img_fnames[7], wbs_img_fnames[8], color=(0,255,0), R=10)
dict_keys(['match_kpts1', 'match_kpts2', 'F', 'num_inl'])
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Close
Hashes for local_feature_tutorial-0.0.3.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | 62d0be66672da33fe92d5fb2d8a3cef97ae515b775e42fb6bcc77023741fcd23 |
|
MD5 | 64d674a61171fff4bff6d46c948f3d4e |
|
BLAKE2b-256 | 9fcd6a06a96f8753fc22e9ff69683ad864eae2029b9d05c1561d8266755fba66 |
Close
Hashes for local_feature_tutorial-0.0.3-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 66b6addd9659620f8d80fd31f520d160c38846475148e886c25fe8c807e14dfc |
|
MD5 | a1ba3b978216ee8c4c8c505c37f05220 |
|
BLAKE2b-256 | 37c1ed48ea7588aaadbd2464223168400020467e170b47e7080988bfb7430ab7 |