Added EigenPlaces for retrieval

README.md

@@ -95,7 +95,7 @@ We show in [`pipeline_SfM.ipynb`](https://nbviewer.jupyter.org/github/cvg/Hierar
 
 - Supported local feature extractors: [SuperPoint](https://arxiv.org/abs/1712.07629), [DISK](https://arxiv.org/abs/2006.13566), [D2-Net](https://arxiv.org/abs/1905.03561), [SIFT](https://www.cs.ubc.ca/~lowe/papers/ijcv04.pdf), and [R2D2](https://arxiv.org/abs/1906.06195).
 - Supported feature matchers: [SuperGlue](https://arxiv.org/abs/1911.11763), its faster follow-up [LightGlue](https://github.com/cvg/LightGlue), and nearest neighbor search with ratio test, distance test, and/or mutual check. hloc also supports dense matching with [LoFTR](https://github.com/zju3dv/LoFTR).
-- Supported image retrieval: [NetVLAD](https://arxiv.org/abs/1511.07247), [AP-GeM/DIR](https://github.com/naver/deep-image-retrieval), [OpenIBL](https://github.com/yxgeee/OpenIBL), and [CosPlace](https://github.com/gmberton/CosPlace).
+- Supported image retrieval: [NetVLAD](https://arxiv.org/abs/1511.07247), [AP-GeM/DIR](https://github.com/naver/deep-image-retrieval), [OpenIBL](https://github.com/yxgeee/OpenIBL), [CosPlace](https://github.com/gmberton/CosPlace) and  [EigenPlaces](https://github.com/gmberton/EigenPlaces).
 
 Using NetVLAD for retrieval, we obtain the following best results:
 

hloc/extract_features.py

@@ -135,9 +135,9 @@
         'model': {'name': 'openibl'},
         'preprocessing': {'resize_max': 1024},
     },
-    'cosplace': {
-        'output': 'global-feats-cosplace',
-        'model': {'name': 'cosplace'},
+    'eigenplaces': {
+        'output': 'global-feats-eigenplaces',
+        'model': {'name': 'eigenplaces'},
         'preprocessing': {'resize_max': 1024},
     }
 }

hloc/extractors/cosplace.py → hloc/extractors/eigenplaces.py

@@ -1,17 +1,26 @@
 '''
-Code for loading models trained with CosPlace as a global features extractor
-for geolocalization through image retrieval.
+Code for loading models trained with EigenPlaces (or CosPlace) as a global
+features extractor for geolocalization through image retrieval.
 Multiple models are available with different backbones. Below is a summary of
 models available (backbone : list of available output descriptors
 dimensionality). For example you can use a model based on a ResNet50 with
 descriptors dimensionality 1024.
+
+EigenPlaces trained models:
+    ResNet18:  [     256, 512]
+    ResNet50:  [128, 256, 512, 2048]
+    ResNet101: [128, 256, 512, 2048]
+    VGG16:     [     512]
+
+CosPlace trained models:
     ResNet18:  [32, 64, 128, 256, 512]
     ResNet50:  [32, 64, 128, 256, 512, 1024, 2048]
     ResNet101: [32, 64, 128, 256, 512, 1024, 2048]
     ResNet152: [32, 64, 128, 256, 512, 1024, 2048]
     VGG16:     [    64, 128, 256, 512]
 
-CosPlace paper: https://arxiv.org/abs/2204.02287
+EigenPlaces paper (ICCV 2023): https://arxiv.org/abs/2308.10832
+CosPlace paper (CVPR 2022): https://arxiv.org/abs/2204.02287
 '''
 
 import torch
@@ -20,15 +29,16 @@
 from ..utils.base_model import BaseModel
 
 
-class CosPlace(BaseModel):
+class EigenPlaces(BaseModel):
     default_conf = {
-        'backbone': 'ResNet50',
+        'variant': 'EigenPlaces',
+        'backbone': 'ResNet101',
         'fc_output_dim' : 2048
     }
     required_inputs = ['image']
     def _init(self, conf):
         self.net = torch.hub.load(
-            'gmberton/CosPlace',
+            'gmberton/' + conf['variant'],
             'get_trained_model',
             backbone=conf['backbone'],
             fc_output_dim=conf['fc_output_dim']