Multiview matching with deep-learning and hand-crafted local features for COLMAP and other SfM software. Supports high-resolution formats and images with rotations. Both CLI and GUI are supported.

Hierarchical Deep Stereo Matching on High Resolution Images, CVPR 2019.

Python (Pytorch) and Matlab (MatConvNet) implementations of CVPR 2021 Image Matching Workshop paper DFM: A Performance Baseline for Deep Feature Matching

Source code and datadset for "Deep learning algorithm for feature matching of cross modality remote sensing images"

Context Encoding for Semantic Segmentation MegaDepth: Learning Single-View Depth Prediction from Internet Photos LiteFlowNet: A Lightweight Convolutional Neural Network for Optical Flow Estimation PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume On the Robustness of Semantic Segmentation Models to Adversarial Attacks SPLATNet: Sparse Lattice Networks for Point Cloud Processing Left-Right Comparative Recurrent Model for Stereo Matching Enhancing the Spatial Resolution of Stereo Images using a Parallax Prior Unsupervised CCA Discovering Point Lights with Intensity Distance Fields CBMV: A Coalesced Bidirectional Matching Volume for Disparity Estimation Learning a Discriminative Feature Network for Semantic Segmentation Revisiting Dilated Convolution: A Simple Approach for Weakly- and Semi- Supervised Semantic Segmentation Unsupervised Deep Generative Adversarial Hashing Network Monocular Relative Depth Perception with Web Stereo Data Supervision Single Image Reflection Separation with Perceptual Losses Zoom and Learn: Generalizing Deep Stereo Matching to Novel Domains EPINET: A Fully-Convolutional Neural Network for Light Field Depth Estimation by Using Epipolar Geometry FoldingNet: Interpretable Unsupervised Learning on 3D Point Clouds Decorrelated Batch Normalization Unsupervised Learning of Depth and Egomotion from Monocular Video Using 3D Geometric Constraints PU-Net: Point Cloud Upsampling Network Real-Time Monocular Depth Estimation using Synthetic Data with Domain Adaptation via Image Style Transfer Tell Me Where To Look: Guided Attention Inference Network Residual Dense Network for Image Super-Resolution Reflection Removal for Large-Scale 3D Point Clouds PlaneNet: Piece-wise Planar Reconstruction from a Single RGB Image Fully Convolutional Adaptation Networks for Semantic Segmentation CRRN: Multi-Scale Guided Concurrent Reflection Removal Network DenseASPP: Densely Connected Networks for Semantic Segmentation SGAN: An Alternative Training of Generative Adversarial Networks Multi-Agent Diverse Generative Adversarial Networks Robust Depth Estimation from Auto Bracketed Images AdaDepth: Unsupervised Content Congruent Adaptation for Depth Estimation DeepMVS: Learning Multi-View Stereopsis GeoNet: Unsupervised Learning of Dense Depth, Optical Flow and Camera Pose GeoNet: Geometric Neural Network for Joint Depth and Surface Normal Estimation Single-Image Depth Estimation Based on Fourier Domain Analysis Single View Stereo Matching Pyramid Stereo Matching Network A Unifying Contrast Maximization Framework for Event Cameras, with Applications to Motion, Depth, and Optical Flow Estimation Image Correction via Deep Reciprocating HDR Transformation Occlusion Aware Unsupervised Learning of Optical Flow PAD-Net: Multi-Tasks Guided Prediciton-and-Distillation Network for Simultaneous Depth Estimation and Scene Parsing Surface Networks Structured Attention Guided Convolutional Neural Fields for Monocular Depth Estimation TextureGAN: Controlling Deep Image Synthesis with Texture Patches Aperture Supervision for Monocular Depth Estimation Two-Stream Convolutional Networks for Dynamic Texture Synthesis Unsupervised Learning of Single View Depth Estimation and Visual Odometry with Deep Feature Reconstruction Left/Right Asymmetric Layer Skippable Networks Learning to See in the Dark

An indoor real scene stereo dataset. It contains 2000 pairs of images with high accuracy disparity maps. We hope it can improve the the generalization performance of deep stereo matching networks.

This project focuses on development of an algorithm for Template Matching on aerial images by implementing classical Computer Vision based techniques and deep-learning based techniques.

paper：基于深度生成匹配网络的光学和SAR图像配准方法研究（Deep Generative Matching Network for Optical and SAR Image Registration）

Official Implementation of Deep Aerial Image Matching using PyTorch

Deep Cross-Modal Projection Learning for Image-Text Matching

A Framework for Deep Learning-based Sparse SAR-Optical Image Matching

MIT-Princeton Vision Toolbox for Robotic Pick-and-Place at the Amazon Robotics Challenge 2017 - Grasp Detection and Image Matching for Novel Objects with Deep Learning

[NeurIPS 2021] TransMatcher: Deep Image Matching Through Transformers for Generalizable Person Re-identification

Deep learning based image denoising using tensorflow/Keras combined with block matching

Deep Adaptive Filtering (DAF) Stereo Networks, DAF-StereoNets for short, leveraging image context as a signal to dynamically guide the matching process

Dataset for Deep Template Matching for Aerial Images

The code of Ada-MVS for "Deep learning based multi-view stereo matching and 3D scene reconstruction from oblique aerial images (ISPRS) "

Interactive Semi Automatic Image 2D Bounding Box Annotation and Labelling Tool using Multi Template Matching An Interactive Semi Automatic Image 2D Bounding Box Annotation/Labelling Tool to aid the Annotater/User to rapidly create 2D Bounding Box Single Object Detection masks for large number of training images in a semi automatic manner in order to train an object detection deep neural network such as Mask R-CNN or U-Net. As the Annotater/User starts annotating/labelling by drawing a bounding box for a few number of images in the selected folder then the algorithm suggests bounding box predictions for the rest of the yet to be annotated/labelled images in the folder. If the predictions are right then the user/annotater can simply press the keyboard key 'y' which indicates that the detected bounding box is correct. If the prediction is wrong then the user/annotater can manually draw a rectangular 2D bounding box over the correct ROI (Region of interest) in the image and then press the key 'y' to proceed further to the rest of the images in the folder. If the user/annotater made a mistake while drawing the 2D bounding box, then he/she can press the key 'n' in order to remove the incorrectly marked 2D bounding box and he/she can repeat the process for the same image until he/she draws the correct 2D bounding box and then after drawing the correct 2D bounding box, the user/annotater may press the key 'y' to continue to the rest of the images. The 2D bounding box prediction over the whole image data set improves as the user/annotater annotates/labels more number of images by drawing 2D bounding boxes. This tool allows the user/annotater to not only interactively and rapidly annotate large number of images but also to validate the predictions at the same time interactively. This tool helps the user/annotater to save a lot of time when annotating/labelling and validating the predictions for a large number of training images in a folder. Instructions to use:- 1. If the training images are in JPEG or any other format, then convert them to PNG format using some other tool or program before using these images for annotation. 2. All the training images must contain the object of interest which is to be annotated. 3. Currently the application only supports 2D bounding box annotation for single object detection per image, but in the future semantic segmentation based annotation features will be added which will allow precise boundary segmentation masks of an object in an image. 4. If some or all of the training images have varying dimensions(shapes/resolutions), then resize them to the same dimensions using this tool by providing the height and width to which all the training images need to be resized to. The height and width are inputed separately in two different dialog boxes which pop up once the program is executed. If the training images need not be resized then press the cancel button in the dialog boxes requesting the height and width. 5. Select the folder containing the training images by navigating to the folder containing the training images through a dialog box which pops up after the program is executed. If the images need to be resized then two dialog boxes pop up. The first dialog box is to navigate to the destination folder containing the unresized raw training images and after resizing another dialog box pops up to navigate to the folder containing the saved resized training images named as "resized_data". If the images need not be resized then only one dialog box pops up so that the user can navigate to the raw training images folder directly. 6. The images in the folder pop up one by one. After drawing the correct 2D bounding box over the ROI (region of Interest), press the 'y' key. Except the first image, the rest of the images will have a 2D bounding box drawn over them. If the predicted box is accurate, then continue by pressing the 'y' key. If the prediction is incorrect, then draw the accurate bounding box and press the 'y' key. If any mistake occured while drawing the 2D box, then reset the image by removing the incorrect drawing by pressing the 'n' key and then draw the correct box and press the 'y' key. 7. The output images are stored in four different folders in the same directory containing the training images folder. among the four folders, one contains the cropped templates of the bounding boxes, black and white mask images, training images and the images with 2D box detection markings.

Because the traditional matching cost calculation algorithm is difficult to meet the accuracy requirements of complex tasks. In recent years, deep learning methods have made successful breakthroughs in image processing. Because of their superior performance in self-learning and feature information extraction, we aim to use deep learning algorithms to calculate high-precision initial matches cost.

This repository contains an implementation of panorama stitching, a computer vision technique used to combine multiple images into a seamless panoramic image. The project leverages classical techniques such as feature detection, matching, and RANSAC, along with a deep learning approach using Homography Net and Tensor DLT.

official repository of paper "An End-to-end Deep Learning Approach for Landmark Detection and Matching in Medical Images"

No description available

The Hybrid Image Matching (HIM) method that combines the deep learning approach with the feature point matching to image classification.

[CVPR'22 Image Matching Workshop (IMW) Top6%] Propose a model registering two images from diverse viewpoints with coarse-to-fine attention, use LoFTR for local feature matching and DKM for regression with deep kernels and Gaussian processes.

Deep Electron Microscopy Image Stitching (DEMIS): a tool for stitching grids of electron microscopy images. Image matching is performed using LoFTR.

This repository provides the official implementation of the paper *Deep learning based multi-view stereo matching and 3D scene reconstruction from oblique aerial images*.

[TIP2024] The code of “Deep Boosting Learning: A Brand-new Cooperative Approach for Image-Text Matching”

UDTATIS is an improved unsupervised deep image stitching system that combines the UDIS++ framework with EfficientLOFTR's feature extraction and matching capabilities. The system is specifically optimized for low-resolution images.

Robust Deep Feature Matching for Multi-modal Remote Sensing Images

The main challenges comprise of classifying the products into right categories and matching the exact same products across different retailers for price comparison. This problem comes under the category of multi-label classification and product retrieval. Sometimes product's textual data (product title & description) is helpful for this task, and sometimes product's images help. Hence, to classify the products into the right category build a Deep Learning multi-modal which incorporates images as well as text for the classification of the Product.