RF-Net: An End-To-End Image Matching Network Based on Receptive Field

Abstract

This paper proposes a new end-to-end trainable matching network based on receptive field, RF-Net, to compute sparse correspondence between images. Building end-toend trainable matching framework is desirable and challenging. The very recent approach, LF-Net, successfully embeds the entire feature extraction pipeline into a jointly trainable pipeline, and produces the state-of-the-art matching results. This paper introduces two modifications to the structure of LF-Net. First, we propose to construct receptive feature maps, which lead to more effective keypoint detection. Second, we introduce a general loss function term, neighbor mask, to facilitate training patch selection. This results in improved stability in descriptor training. We trained RF-Net on the open dataset HPatches, and compared it with other methods on multiple benchmark datasets. Experiments show that RF-Net outperforms existing state-of-the-art methods. * Corresponding author.to make them optimally cooperate with each other, hence, is more desirable. However, training such a network is difficult because the two subcomponents have their individually different objectives to optimize. Not many successful end-to-end matching pipelines have been reported in literatures. LIFT [29] is probably the first notable design towards this goal. However, LIFT relies on the output of SIFT detector to initialize the training, and hence, its detector behaves similarly to the SIFT detector. The recent network, SuperPoint [5], achieves this end-to-end training. But its detector needs to be pre-trained on synthetic image sets, and whole network is trained using images under synthesized affine transformations. The more recent LF-Net [18] is inspired by Q-learning, and uses a Siamese architecture to train the entire network without the help of any hand-craft method. In this paper, we develop an end-to-end matching network with enhanced detector and descriptor training modules, which we elaborate as follows. Feature Maps shared 5 × 5

Wire Structure Image-Based 3d Reconstruction Aided by Deep Learning

“…Feature matching networks (Yi et al, 2016, Ono et al, 2018, Christiansen et al, 2019, Shen et al, 2019, Kniaz et al, 2020 seems to outperform handcrafted feature detectors/descriptor methods. Still, their performance is closely related to the similarity of local image patches in the training dataset with respect to the images used during inference.…”

Registration of Multiresolution Remote Sensing Images Based on L2-Siamese Model

“…The experiments were performed using different remote sensing images to evaluate the performance and robustness of the algorithm. Seven remote sensing registration algorithms were used as comparison groups: SIFT [14], SURF [16], FSC-SIFT [20], PSO-SIFT [17], SAR-SIFT [19], PSO-SIFT-CNN [23] and RF-Net [31].…”

“…The fourth approach is based on deep network image matching (registration) not only can extract the deep features of key points, but also can automatically extract key points.Methods that apply this approach include D2-Net [29], Super-Point [30], and RF-Net models [31]. These methods are commonly used for natural image matching, and have good robustness to angle and illumination.…”

MSU-Net: Multi-Scale U-Net for 2D Medical Image Segmentation

“…For example, in the pulmonary lesion or multi-organ segmentation task, the edge detail of the smaller lesion/organ is not fine by the large receptor field and the structure of the lesion/organ is not obvious by the small receptor field. Therefore, it is very important to use the convolution kernel with different receptive fields to process the image (Luo et al, 2016 ; Peng et al, 2017 ; Shen et al, 2019 ). In the natural image processing task, satisfactory results are obtained by combining the convolution of different receptive fields (Seif and Androutsos, 2018 ).…”