Skip to main content

No project description provided

Project description

easy-local-features-baselines

Just some scripts to make things easier for the local features baselines.

Installation

# make sure you have torch installed
# pip install torch torchvision
pip install easy-local-features

Installing from source

You may want to install from source if you want to modify the code or if you want to use the latest version. To do so, you can clone this repository and install the requirements.

I suggest using a conda environment to install the requirements. You can create one using the following command.

conda create -n elf python=3.9 # the python version is not so critical, but I used 3.9.
conda activate elf

Now we can install everything.

pip install -r requirements.txt
pip install -e .

How to use

# Choose you extractor
from easy_local_features.feature.baseline_superpoint import SuperPoint_baseline
# from easy_local_features.feature.baseline_deal import DEAL_baseline
# from easy_local_features.feature.baseline_dalf import DALF_baseline
# from easy_local_features.feature.baseline_aliked import ALIKED_baseline
# from easy_local_features.feature.baseline_alike import ALIKE_baseline
# from easy_local_features.feature.baseline_disk import DISK_baseline
# from easy_local_features.feature.baseline_r2d2 import R2D2_baseline

# also a matcher
from easy_local_features.matching.baseline_lightglue import LightGlue_baseline
# from easy_local_features.matching.baseline_superglue import SuperGlue_baseline
# from easy_local_features.matching.baseline_loftr import LoFTR_baseline
import cv2

# Load an image
img = cv2.imread("assets/notredame.png")

# Initialize the extractor
extractor = SuperPoint_baseline()
matcher = LightGlue_baseline() # works with superpoint and disk

# Return keypoints and descriptors just like OpenCV
keypoints0, descriptors0 = extractor.detectAndCompute(img)
keypoints1, descriptors1 = extractor.detectAndCompute(img)

# Match the descriptors
mkpts0, mkpts1, matches = matcher.match(keypoints0, keypoints1, descriptors0, descriptors1)

img = cv2.drawMatches(img, keypoints0, img, keypoints1, matches, None, flags=cv2.DrawMatchesFlags_NOT_DRAW_SINGLE_POINTS)
cv2.imshow("Matched", img)
cv2.waitKey(0)

TODO

  • Add a setup.py to make it a pip package
  • Make a general maching class
    • The idea is to have a class that can match images using any local feature extractor and any matching method.
  • Fix requirements to install automatically with the package (maybe)
  • Add a script to download some datasets
  • Add more baselines :)
    • DEAL
    • DALF
    • DKM
    • ASLFeat
    • R2D2
    • DISK
    • SuperPoint
    • SuperGlue
    • LightGlue
    • LoFTR
    • ALIKE
    • ALIKED
      • Add LICENSE file
      • Test on MAC M1 with CPU
      • Test on Linux with CPU
      • Test with CUDA

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

easy_local_features-0.3.12.tar.gz (122.3 kB view details)

Uploaded Source

File details

Details for the file easy_local_features-0.3.12.tar.gz.

File metadata

  • Download URL: easy_local_features-0.3.12.tar.gz
  • Upload date:
  • Size: 122.3 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.10.13

File hashes

Hashes for easy_local_features-0.3.12.tar.gz
Algorithm Hash digest
SHA256 eca63e200ee8961907d6aae9fb27c2bfbaf43a69256cefc23d7be97d0a30d040
MD5 777108842942af0836eb5ae97e212534
BLAKE2b-256 44014191f808f13e11240770dbb8149c741760dc776b876bf3c32e66fa562f7e

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page