Code release.

.gitignore

@@ -1,3 +1,36 @@
+# Prerequisites
+*.d
+
+# Compiled Object files
+*.slo
+*.lo
+*.o
+*.obj
+
+# Precompiled Headers
+*.gch
+*.pch
+
+# Compiled Dynamic libraries
+*.so
+*.dylib
+*.dll
+
+# Fortran module files
+*.mod
+*.smod
+
+# Compiled Static libraries
+*.lai
+*.la
+*.a
+*.lib
+
+# Executables
+*.exe
+*.out
+*.app
+
 # Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]

Dockerfile

@@ -0,0 +1,45 @@
+# Base image
+FROM tensorflow/tensorflow:1.15.5-py3
+
+# Install necessary dependencies
+RUN apt-get update && apt-get install -y \
+    cmake \
+    pkg-config \
+    wget \
+    unzip && \
+    rm -rf /var/lib/apt/lists/*
+
+# Set working directory
+WORKDIR /src
+
+#Install OpenCV 3.4.8
+RUN wget "https://github.com/opencv/opencv/archive/3.4.8.zip" && \
+    unzip 3.4.8.zip && \
+    mkdir -p opencv-3.4.8/build && \
+    cd opencv-3.4.8/build && \
+    cmake -DCMAKE_BUILD_TYPE=Release \
+          -DBUILD_LIST=core,highgui,imgproc,features2d .. && \
+    make -j$(nproc) && \
+    make install
+
+
+# Download and extract SimulationICCV.tar.gz
+RUN wget "https://www.verlab.dcc.ufmg.br/nonrigid/SimulationICCV.tar.gz" && \
+    tar -xvf SimulationICCV.tar.gz
+
+RUN pip3 install opencv-python-headless==3.4.3.18 scipy --prefer-binary
+
+# Copy geopatch source from the current directory into the container
+COPY . /src
+
+# Build geopatch-forge/geopatch
+RUN cd /src/geopatch && \
+    mkdir -p build && cd build && \
+    cmake .. && make -j$(nproc)
+
+# Clean up temporary files to reduce image size
+RUN rm -rf /src/opencv-3.4.8 /src/3.4.8.zip /src/SimulationICCV.tar.gz
+
+# Default command
+CMD ["/bin/bash"]
+

README.md

@@ -0,0 +1,93 @@
+[![License](https://img.shields.io/badge/License-GPLv3-blue.svg)](https://www.gnu.org/licenses/gpl-3.0)
+
+
+# GeoPatch Descriptor
+
+This repository contains the original implementation of the descriptor "<b>Learning Geodesic-Aware Local Features from RGB-D Images</b>", published at Computer Vision and Image Understanding journal. GeoPatch is by design invariant to isometric non-rigid deformations of surfaces by leveraging geodesic-invariant sampling, designed as a mapping function before descriptor computation is performed from photometric information.
+
+
+## <b>Learning Geodesic-Aware Local Features from RGB-D Images</b> <br>[[Project Page]](https://www.verlab.dcc.ufmg.br/descriptors/cviu2022/) [[Paper]](https://arxiv.org/abs/2203.12016)
+
+<img src='./images/teaser.jpg' align="center" width=900 />
+
+
+
+## Ready to use Docker container
+
+We recommend running the project with Docker, which requires a single command to build the entire project.
+
+First, build the container:
+
+```sh
+docker compose build
+```
+
+Then, run the docker in interactive mode:
+
+```sh
+docker compose run --rm geopatch
+```
+
+Finally, you can run the provided demo, which runs both geodesic patch extraction and local feature computation:
+
+```sh
+sh run_demo.sh
+```
+
+Notice that the output files are being saved inside the container.
+
+## Ready to use Singularity container
+
+Alternatively we also provide a singularity recipe so you can easily and smoothly build the project.
+
+First, build the container:
+
+```sh
+sudo singularity build geopatch.sif Singularity.geopatch
+```
+
+Then, run the container in interactive mode:
+
+```sh
+singularity shell --writable-tmpfs --pwd /src geopatch.sif
+```
+
+Finally, you can run the provided demo, which runs both geodesic patch extraction and local feature computation:
+
+```sh
+sh run_demo.sh
+```
+
+Notice that the output files are being saved inside the container.
+
+## Datasets
+
+All available datasets and ground-truth files are available for download at [https://verlab.dcc.ufmg.br/descriptors](https://verlab.dcc.ufmg.br/descriptors)
+
+
+## Non-Rigid Simulator
+
+The code for the non-rigid simulator used in our work is available in the [`nonrigid_sim`](nonrigid_sim/) folder. For detailed instructions on usage, please refer to the [`README.md`](nonrigid_sim/README.md) inside the folder.
+
+
+## Published works
+
+If you find this code useful for your research, please cite the paper:
+
+```
+@article{potje2022learning,
+  title={Learning geodesic-aware local features from RGB-D images},
+  author={Potje, Guilherme and Martins, Renato and Cadar, Felipe and Nascimento, Erickson R},
+  journal={Computer Vision and Image Understanding},
+  volume={219},
+  pages={103409},
+  year={2022},
+  publisher={Elsevier}
+}
+```
+
+**VeRLab:** Laboratory of Computer Vison and Robotics https://www.verlab.dcc.ufmg.br
+<br>
+<img align="left" width="auto" height="50" src="./images/ufmg.png">
+<img align="right" width="auto" height="50" src="./images/verlab.png">
+<br/>

Singularity.geopatch

@@ -0,0 +1,45 @@
+Bootstrap: docker
+From: tensorflow/tensorflow:1.15.5-py3
+
+%files
+    . /src
+
+%post
+    # Install necessary dependencies
+    apt-get update && apt-get install -y \
+        cmake \
+        pkg-config \
+        wget \
+        unzip && \
+        rm -rf /var/lib/apt/lists/*
+
+    # Set working directory
+    cd /src
+
+    # Install OpenCV 3.4.8
+    wget "https://github.com/opencv/opencv/archive/3.4.8.zip" && \
+    unzip 3.4.8.zip && \
+    mkdir -p opencv-3.4.8/build && \
+    cd opencv-3.4.8/build && \
+    cmake -DCMAKE_BUILD_TYPE=Release \
+          -DBUILD_LIST=core,highgui,imgproc,features2d .. && \
+    make -j$(nproc) && \
+    make install
+
+    # Download and extract SimulationICCV.tar.gz
+    cd /src
+    wget "https://www.verlab.dcc.ufmg.br/nonrigid/SimulationICCV.tar.gz" && \
+    tar -xvf SimulationICCV.tar.gz
+
+    # Install Python dependencies
+    pip3 install opencv-python-headless==3.4.3.18 scipy --prefer-binary
+
+    # Build geopatch
+    cd /src/geopatch
+    mkdir -p build && cd build
+    cmake .. && make -j$(nproc)
+
+    # Clean up temporary files
+    rm -rf /src/opencv-3.4.8 /src/3.4.8.zip /src/SimulationICCV.tar.gz
+
+    chmod -R 777 /src

architecture.py

@@ -0,0 +1,61 @@
+
+#%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+#%   This file is part of https://github.com/verlab/GeoPatch_CVIU_2022
+#
+#   geopatch-descriptor is free software: you can redistribute it and/or modify
+#   it under the terms of the GNU General Public License as published by
+#   the Free Software Foundation, either version 3 of the License, or
+#   (at your option) any later version.
+#
+#   geopatch-descriptor is distributed in the hope that it will be useful,
+#   but WITHOUT ANY WARRANTY; without even the implied warranty of
+#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+#   GNU General Public License for more details.
+#
+#   You should have received a copy of the GNU General Public License
+#   along with geopatch-descriptor.  If not, see <http://www.gnu.org/licenses/>.
+#%
+#%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+
+from __future__ import absolute_import
+from __future__ import print_function
+import numpy as np
+import cv2
+import math
+
+from tensorflow.keras.utils import plot_model
+from tensorflow.keras.models import Model
+from tensorflow.keras import backend as K
+from tensorflow.keras.models import model_from_json
+
+import os
+###############################################################
+#Configure Session for Memory Limit on TensorFlow using Keras
+# session config
+import tensorflow as tf
+config = tf.ConfigProto()
+config.gpu_options.allow_growth = True
+tf.keras.backend.set_session(tf.Session(config=config))
+###############################################################
+
+pretrained_path = os.path.dirname(os.path.realpath(__file__))  +'/weights'
+
+class TinyDesc(object):
+	model = None
+	PATCH_SIZE = 32
+
+	def __init__(self, model_name):
+		# load json and create model
+		json_file = open(pretrained_path + '/' + model_name + '.json', 'r')
+		loaded_model_json = json_file.read()
+		json_file.close()
+		loaded_model = model_from_json(loaded_model_json)
+		# load weights into new model
+		loaded_model.load_weights(pretrained_path + '/' + model_name + '.h5')
+
+		print("Loaded model from disk")
+
+		# evaluate loaded model on test data
+		#loaded_model.compile(loss='mean_squared_error', optimizer='rmsprop', metrics=['accuracy'])
+
+		self.model = loaded_model

docker-compose.yml

@@ -0,0 +1,10 @@
+version: '3.8'
+
+services:
+  geopatch:
+    image: geopatch-1.0  
+    build: .            
+    container_name: geopatch_container
+    stdin_open: true
+    tty: true
+