✨ Add RIPE feature extractor support and submodule integration (#136)

📝 Update README and config files to include RIPE algorithm
🔧 Fix github URL typos in config.yaml for RDD entries

Author: Vincentqyw

.gitmodules

@@ -82,3 +82,6 @@
 [submodule "imcui/third_party/rdd"]
     path = imcui/third_party/rdd
     url = https://github.com/agipro/rdd.git
+[submodule "imcui/third_party/RIPE"]
+    path = imcui/third_party/RIPE
+    url = https://github.com/fraunhoferhhi/RIPE

README.md

@@ -33,6 +33,7 @@ The tool currently supports various popular image matching algorithms, namely:
 
 | Algorithm        | Supported | Conference/Journal | Year | GitHub Link |
 |------------------|-----------|--------------------|------|-------------|
+| RIPE           | ✅ | ICCV    | 2025 | [Link](https://github.com/fraunhoferhhi/RIPE)  |
 | RDD            | ✅ | CVPR    | 2025 | [Link](https://github.com/xtcpete/rdd)  |
 | LiftFeat       | ✅ | ICRA    | 2025 | [Link](https://github.com/lyp-deeplearning/LiftFeat) |
 | DaD            | ✅ | ARXIV   | 2025 | [Link](https://github.com/Parskatt/dad) |

config/config.yaml

@@ -253,14 +253,25 @@ matcher_zoo:
       paper: https://arxiv.org/abs/2505.0342
       project: null
       display: true
+  ripe(+mnn):
+    matcher: NN-mutual
+    feature: ripe
+    dense: false
+    info:
+      name: RIPE #dispaly name
+      source: "ICCV 2025"
+      github: https://github.com/fraunhoferhhi/RIPE
+      paper: https://arxiv.org/abs/2507.04839
+      project: https://fraunhoferhhi.github.io/RIPE
+      display: true
   rdd(sparse):
     matcher: NN-mutual
     feature: rdd
     dense: false
     info:
       name: RDD(sparse) #dispaly name
       source: "CVPR 2025"
-      github: hhttps://github.com/xtcpete/rdd
+      github: https://github.com/xtcpete/rdd
       paper: https://arxiv.org/abs/2505.08013
       project: https://xtcpete.github.io/rdd
       display: true
@@ -270,7 +281,7 @@ matcher_zoo:
     info:
       name: RDD(dense) #dispaly name
       source: "CVPR 2025"
-      github: hhttps://github.com/xtcpete/rdd
+      github: https://github.com/xtcpete/rdd
       paper: https://arxiv.org/abs/2505.08013
       project: https://xtcpete.github.io/rdd
       display: true

imcui/hloc/extract_features.py

@@ -236,6 +236,17 @@
             "resize_max": 1600,
         },
     },
+    "ripe": {
+        "output": "feats-ripe-n2048-r1600",
+        "model": {
+            "name": "ripe",
+            "max_keypoints": 2048,
+        },
+        "preprocessing": {
+            "grayscale": False,
+            "resize_max": 1600,
+        },
+    },
     "aliked-n16-rot": {
         "output": "feats-aliked-n16-rot",
         "model": {

imcui/hloc/extractors/ripe.py

@@ -0,0 +1,46 @@
+import sys
+from pathlib import Path
+from ..utils.base_model import BaseModel
+from .. import logger, MODEL_REPO_ID
+
+ripe_path = Path(__file__).parent / "../../third_party/RIPE"
+sys.path.append(str(ripe_path))
+
+from ripe import vgg_hyper
+
+
+class RIPE(BaseModel):
+    default_conf = {
+        "keypoint_threshold": 0.05,
+        "max_keypoints": 5000,
+        "model_name": "weights_ripe.pth",
+    }
+
+    required_inputs = ["image"]
+
+    def _init(self, conf):
+        logger.info("Loading RIPE model...")
+        model_path = self._download_model(
+            repo_id=MODEL_REPO_ID,
+            filename="{}/{}".format(Path(__file__).stem, self.conf["model_name"]),
+        )
+        self.net = vgg_hyper(Path(model_path))
+        logger.info("Loading RIPE model done!")
+
+    def _forward(self, data):
+        keypoints, descriptors, scores = self.net.detectAndCompute(
+            data["image"], threshold=0.5, top_k=2048
+        )
+
+        if self.conf["max_keypoints"] < len(keypoints):
+            idxs = scores.argsort()[-self.conf["max_keypoints"] or None :]
+            keypoints = keypoints[idxs, :2]
+            descriptors = descriptors[idxs]
+            scores = scores[idxs]
+
+        pred = {
+            "keypoints": keypoints[None],
+            "descriptors": descriptors[None].permute(0, 2, 1),
+            "scores": scores[None],
+        }
+        return pred