Skip to main content

Add your description here

Project description

easy-local-features

Unified, minimal wrappers around many local feature extractors and matchers (classical + learned).

⚠️ CRITICAL LICENSE DISCLAIMER

This repository aggregates wrappers around MANY third‑party local feature extractors and matchers. Each baseline/model keeps its OWN original license (BSD, MIT, Apache 2.0, GPLv3, Non‑Commercial, CC BY‑NC‑SA, custom research licenses, etc.). Your rights for a given baseline are governed only by that baseline’s upstream license. Some components here are non‑commercial only (e.g., SuperGlue, original SuperPoint, R2D2) or copyleft (e.g., DISK under GPLv3). Others are permissive (BSD/MIT/Apache 2.0). Before any research publication, internal deployment, redistribution, or commercial/production use, YOU MUST review and comply with every relevant upstream license, including attribution, notice reproduction, share‑alike, copyleft, and patent clauses.

The maintainers provide NO warranty, NO guarantee of license correctness, and accept NO liability for misuse. This notice and any summaries are not legal advice. If in doubt, consult qualified counsel.

See: LICENSES.md for an overview and links to included full license texts. Built with DINOv3.

Installation

pip install easy-local-features

Installing from source

pip install -e .

Usage

Stable minimal API

  • getExtractor(name, conf) → returns an extractor
  • .to(device)"cpu" | "cuda" | "mps"
  • .match(img0, img1){"mkpts0": (M,2), "mkpts1": (M,2), ...}
  • Descriptor-only methods additionally support .addDetector(detector)

Detector-only methods

  • Some methods only implement detect(image) -> keypoints (no descriptors, no matching). Use getDetector(name, conf) for those.

Example:

from easy_local_features import getDetector
from easy_local_features.utils import io

img = io.fromPath("tests/assets/megadepth0.jpg")
det = getDetector("rekd", {"num_keypoints": 1500}).to("cpu")
kps = det.detect(img)  # [1, N, 2]

Detect+Describe (one-liner matching)

from easy_local_features import getExtractor
from easy_local_features.utils import io, ops, vis

img0 = io.fromPath("tests/assets/megadepth0.jpg")
img1 = io.fromPath("tests/assets/megadepth1.jpg")
img0 = ops.resize_short_edge(img0, 320)[0]
img1 = ops.resize_short_edge(img1, 320)[0]

extractor = getExtractor("aliked", {"top_k": 2048}).to("cpu")
out = extractor.match(img0, img1)

vis.plot_pair(img0, img1, title="ALIKED")
vis.plot_matches(out["mkpts0"], out["mkpts1"])
vis.save("tests/results/aliked.png")

Descriptor-only (attach a detector)

from easy_local_features import getExtractor
from easy_local_features.feature.baseline_superpoint import SuperPoint_baseline
from easy_local_features.utils import io, ops

img0 = ops.resize_short_edge(io.fromPath("tests/assets/megadepth0.jpg"), 320)[0]
img1 = ops.resize_short_edge(io.fromPath("tests/assets/megadepth1.jpg"), 320)[0]

desc = getExtractor("sosnet").to("cpu")
det = SuperPoint_baseline({"top_k": 2048, "detection_threshold": 0.005}).to("cpu")
desc.addDetector(det)

out = desc.match(img0, img1)

End-to-end matchers

from easy_local_features import getExtractor
from easy_local_features.utils import io, ops

img0 = ops.resize_short_edge(io.fromPath("tests/assets/megadepth0.jpg"), 320)[0]
img1 = ops.resize_short_edge(io.fromPath("tests/assets/megadepth1.jpg"), 320)[0]

roma = getExtractor("romav2", {"top_k": 2000}).to("cpu")
out = roma.match(img0, img1)

CLI

easy-local-features --list
easy-local-features --list-detectors

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

easy_local_features-0.8.18.tar.gz (4.5 MB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

easy_local_features-0.8.18-py3-none-any.whl (506.6 kB view details)

Uploaded Python 3

File details

Details for the file easy_local_features-0.8.18.tar.gz.

File metadata

File hashes

Hashes for easy_local_features-0.8.18.tar.gz
Algorithm Hash digest
SHA256 756a77a29e63c1e68083f285096d33bf80140c4bf1cb92e7322a81b6cb8023ce
MD5 0b570020fe709bf71340ab43b8111996
BLAKE2b-256 1be8d5e131489f83a8d11f78fdfb5ca78c0a7e58b0e82b1397472e6ad9842ddb

See more details on using hashes here.

File details

Details for the file easy_local_features-0.8.18-py3-none-any.whl.

File metadata

File hashes

Hashes for easy_local_features-0.8.18-py3-none-any.whl
Algorithm Hash digest
SHA256 7d660607a91e84b532dfd16d4c14838c7630ba920983bcea4d1462cd7c2ad46e
MD5 5a863be5dfef01dc6392a775d64f558c
BLAKE2b-256 c5156316b770ab457433c3e8bfc898079edcb3c5612ceb14f0ab9595f1d3eb9f

See more details on using hashes here.

Supported by

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