search

mihaidusmanu / d2-net

D2-Net: A Trainable CNN for Joint Description and Detection of Local Features
Other
761 stars 163 forks source link

HPatches evaluation by SuperPoint and RootSIFT #74

Closed coper16 closed 3 years ago

coper16 commented 3 years ago

Hi! Thank you for your great work!

I want to use your HPatches evaluation code. but, I cannot understand how to apply SuperPointNet and RootSIFT to this. (for example, SuperPointNet outputs Keypoint and descriptor. But, d2-net(model_test.py) outputs dense_feature,detections,and displacement.)

Could you tell me how to apply SuperPoint and RootSIFT to feature extraction code(extract_features.py,model_test.py and pyramid.py)?

mihaidusmanu commented 3 years ago

You should not use model_test.py since that is only an internal class. You should use extract_features.py which directly returns keypoints and descriptors as explained at the following link https://github.com/mihaidusmanu/d2-net#feature-extraction.

For the HPatches code, there are instructions on running it in the associated folder: https://github.com/mihaidusmanu/d2-net/tree/master/hpatches_sequences#hpatches-sequences--image-pairs-matching-benchmark.

There is no code for SuperPoint in this repo, please refer to the official repository.