This paper addresses the domain shift problem for segmentation. As a solution, we propose OLVA, a novel and lightweight unsupervised domain adaptation method based on a Variational Auto-Encoder (VAE) and Optimal Transport (OT) theory. Thanks to the VAE, our model learns a shared cross-domain latent space that follows a normal distribution, which reduces the domain shift. To guarantee valid segmentations, our shared latent space is designed to model the shape rather than the intensity variations. We further rely on an OT loss to match and align the remaining discrepancy between the two domains in the latent space. We demonstrate OLVA's effectiveness for the segmentation of multiple cardiac structures on the public Multi-Modality Whole Heart Segmentation (MM-WHS) dataset, where the source domain consists of annotated 3D MR images and the unlabelled target domain of 3D CTs. Our results show remarkable improvements with an additional margin of 12.5\% dice score over concurrent generative training approaches.
Synthetic data used for scene text detection and recognition tasks have proven effective. However, there are still two problems: First, the color schemes used for text coloring in the existing methods are relatively fixed color key-value pairs learned from real datasets. The dirty data in real datasets may cause the problem that the colors of text and background are too similar to be distinguished from each other. Second, the generated texts are uniformly limited to the same depth of a picture, while there are special cases in the real world that text may appear across depths. To address these problems, in this paper we design a novel method to generate color schemes, which are consistent with the characteristics of human eyes to observe things. The advantages of our method are as follows: (1) overcomes the color confusion problem between text and background caused by dirty data; (2) the texts generated are allowed to appear in most locations of any image, even across depths; (3) avoids analyzing the depth of background, such that the performance of our method exceeds the state-of-the-art methods; (4) the speed of generating images is fast, nearly one picture generated per three milliseconds. The effectiveness of our method is verified on several public datasets.
Precise geolocalization is crucial for unmanned aerial vehicles (UAVs). However, most current deployed UAVs rely on the global navigation satellite systems (GNSS) or high precision inertial navigation systems (INS) for geolocalization. In this paper, we propose to use a lightweight visual-inertial system with a 2D georeference map to obtain accurate and consecutive geodetic positions for UAVs. The proposed system firstly integrates a micro inertial measurement unit (MIMU) and a monocular camera as odometry to consecutively estimate the navigation states and reconstruct the 3D position of the observed visual features in the local world frame. To obtain the geolocation, the visual features tracked by the odometry are further registered to the 2D georeferenced map. While most conventional methods perform image-level aerial image registration, we propose to align the reconstructed points to the map points in the geodetic frame; this helps to filter out the large portion of outliers and decouples the negative effects from the horizontal angles. The registered points are then used to relocalize the vehicle in the geodetic frame. Finally, a pose graph is deployed to fuse the geolocation from the aerial image registration and the local navigation result from the visual-inertial odometry (VIO) to achieve consecutive and drift-free geolocalization performance. We have validated the proposed method by installing the sensors to a UAV body rigidly and have conducted two flights in different environments with unknown initials. The results show that the proposed method can achieve less than 4m position error in flight at 100m high and less than 9m position error in flight about 300m high.
Multi-Focus Image Fusion (MFIF) is one of the promising techniques to obtain all-in-focus images to meet people's visual needs and it is a precondition of other computer vision tasks. One of the research trends of MFIF is to solve the defocus spread effect (DSE) around the focus/defocus boundary (FDB). In this paper, we present a novel generative adversarial network termed MFIF-GAN to translate multi-focus images into focus maps and to get the all-in-focus images further. The Squeeze and Excitation Residual Network (SE-ResNet) module as an attention mechanism is employed in the network. During the training, we propose reconstruction and gradient regularization loss functions to guarantee the accuracy of generated focus maps. In addition, by combining the prior knowledge of training conditon, this network is trained on a synthetic dataset with DSE based on an {\alpha}-matte model. A series of experimental results demonstrate that the MFIF-GAN is superior to several representative state-of-the-art (SOTA) algorithms in visual perception, quantitative analysis as well as efficiency.
In this paper, we present a novel neuroevolutionary method to identify the architecture and hyperparameters of convolutional autoencoders. Remarkably, we used a hypervolume indicator in the context of neural architecture search for autoencoders, for the first time to our current knowledge. Results show that images were compressed by a factor of more than 10, while still retaining enough information to achieve image classification for the majority of the tasks. Thus, this new approach can be used to speed up the AutoML pipeline for image compression.
Currently, the state-of-the-art methods treat few-shot semantic segmentation task as a conditional foreground-background segmentation problem, assuming each class is independent. In this paper, we introduce the concept of meta-class, which is the meta information (e.g. certain middle-level features) shareable among all classes. To explicitly learn meta-class representations in few-shot segmentation task, we propose a novel Meta-class Memory based few-shot segmentation method (MM-Net), where we introduce a set of learnable memory embeddings to memorize the meta-class information during the base class training and transfer to novel classes during the inference stage. Moreover, for the $k$-shot scenario, we propose a novel image quality measurement module to select images from the set of support images. A high-quality class prototype could be obtained with the weighted sum of support image features based on the quality measure. Experiments on both PASCAL-$5^i$ and COCO dataset shows that our proposed method is able to achieve state-of-the-art results in both 1-shot and 5-shot settings. Particularly, our proposed MM-Net achieves 37.5\% mIoU on the COCO dataset in 1-shot setting, which is 5.1\% higher than the previous state-of-the-art.
Indoor scene images usually contain scattered objects and various scene layouts, which make RGB-D scene classification a challenging task. Existing methods still have limitations for classifying scene images with great spatial variability. Thus, how to extract local patch-level features effectively using only image labels is still an open problem for RGB-D scene recognition. In this paper, we propose an efficient framework for RGB-D scene recognition, which adaptively selects important local features to capture the great spatial variability of scene images. Specifically, we design a differentiable local feature selection (DLFS) module, which can extract the appropriate number of key local scenerelated features. Discriminative local theme-level and object-level representations can be selected with the DLFS module from the spatially-correlated multi-modal RGB-D features. We take advantage of the correlation between RGB and depth modalities to provide more cues for selecting local features. To ensure that discriminative local features are selected, the variational mutual information maximization loss is proposed. Additionally, the DLFS module can be easily extended to select local features of different scales. By concatenating the local-orderless and global structured multi-modal features, the proposed framework can achieve state-of-the-art performance on public RGB-D scene recognition datasets.
Image denoising is a well studied problem with an extensive activity that has spread over several decades. Despite the many available denoising algorithms, the quest for simple, powerful and fast denoisers is still an active and vibrant topic of research. Leading classical denoising methods are typically designed to exploit the inner structure in images by modeling local overlapping patches. In contrast, recent newcomers to this arena are supervised neural-network-based methods that bypass this modeling altogether, targeting the inference goal directly and globally, while tending to be very deep and parameter heavy. This work proposes a novel low-weight learnable architecture that embeds in it several of the main concepts from the classical methods, while being trained for best denoising performance. More specifically, our proposed network relies on patch processing, leveraging non-local self-similarity, representation sparsity and a multiscale treatment. The proposed architecture achieves near state-of-the-art denoising results, while using a small fraction of the typical number of parameters. Furthermore, we demonstrate the ability of the proposed network to adapt itself to an incoming image by leveraging similar clean ones.
Hypercomplex neural networks have proved to reduce the overall number of parameters while ensuring valuable performances by leveraging the properties of Clifford algebras. Recently, hypercomplex linear layers have been further improved by involving efficient parameterized Kronecker products. In this paper, we define the parameterization of hypercomplex convolutional layers to develop lightweight and efficient large-scale convolutional models. Our method grasps the convolution rules and the filters organization directly from data without requiring a rigidly predefined domain structure to follow. The proposed approach is flexible to operate in any user-defined or tuned domain, from 1D to $n$D regardless of whether the algebra rules are preset. Such a malleability allows processing multidimensional inputs in their natural domain without annexing further dimensions, as done, instead, in quaternion neural networks for 3D inputs like color images. As a result, the proposed method operates with $1/n$ free parameters as regards its analog in the real domain. We demonstrate the versatility of this approach to multiple domains of application by performing experiments on various image datasets as well as audio datasets in which our method outperforms real and quaternion-valued counterparts.
Social concepts referring to non-physical objects--such as revolution, violence, or friendship--are powerful tools to describe, index, and query the content of visual data, including ever-growing collections of art images from the Cultural Heritage (CH) field. While much progress has been made towards complete image understanding in computer vision, automatic detection of social concepts evoked by images is still a challenge. This is partly due to the well-known semantic gap problem, worsened for social concepts given their lack of unique physical features, and reliance on more unspecific features than concrete concepts. In this paper, we propose the translation of recent cognitive theories about social concept representation into a software approach to represent them as multimodal frames, by integrating multisensory data. Our method focuses on the extraction, analysis, and integration of multimodal features from visual art material tagged with the concepts of interest. We define a conceptual model and present a novel ontology for formally representing social concepts as multimodal frames. Taking the Tate Gallery's collection as an empirical basis, we experiment our method on a corpus of art images to provide a proof of concept of its potential. We discuss further directions of research, and provide all software, data sources, and results.