Pairs Of Patches Research Articles

Unveiling image source: Instance-level camera device linking via context-aware deep Siamese network

Unveiling the source of an image is one of the most effective ways to validate the originality, authenticity, and reliability in the field of digital forensics. Source camera device identification can identify the specific camera device used to take a photo under investigation. While great progress has been made by the photo-response non-uniformity (PRNU)-based methods over the past decade, the challenge of instance-level source camera device linking, which verifies whether two images in question were captured by the same camera device, remains significant. This challenge is mainly due to the absence of auxiliary images to construct a clean camera fingerprint for each camera, particularly dealing with small image sizes. To overcome this limitation, in this paper, we formulate the task of source device linking as a binary classification problem and propose a simple yet effective framework based on a context-aware deep Siamese network. We take advantage of a Siamese architecture to extract the intrinsic camera device-related noise patterns from a pair of image patches in parallel for comparisons without any auxiliary images. Moreover, a recurrent criss-cross group is utilized to aggregate contextual information in the noise residual maps to alleviate the problem that PRNU noise maps are easily contaminated by the additive noises from image contents. For reliable device linking, we employ a patch-selection strategy on a pair of test images to adaptively choose suitable image patch pairs according to image contents. The final decision of a pair of test images is obtained from the average similarity score of the selected image patch pairs. Compared with existing state-of-the-art methods, our proposed framework can achieve better performance on both the tasks of source camera identification and source device linking without any prior knowledge, i.e., reliable camera fingerprints, regardless of whether the camera devices are “seen” or “unseen” in the training stage. The experimental results on two standard image forensic datasets demonstrate that the proposed method not only shows robustness with respect to different image patch sizes and image quality degenerations, but also has a generalization ability across digital camera and smartphone devices.

TCDM: Transformational Complexity Based Distortion Metric for Perceptual Point Cloud Quality Assessment.

The goal of objective point cloud quality assessment (PCQA) research is to develop quantitative metrics that measure point cloud quality in a perceptually consistent manner. Merging the research of cognitive science and intuition of the human visual system (HVS), in this article, we evaluate the point cloud quality by measuring the complexity of transforming the distorted point cloud back to its reference, which in practice can be approximated by the code length of one point cloud when the other is given. For this purpose, we first make space segmentation for the reference and distorted point clouds based on a 3D Voronoi diagram to obtain a series of local patch pairs. Next, inspired by the predictive coding theory, we utilize a space-aware vector autoregressive (SA-VAR) model to encode the geometry and color channels of each reference patch with and without the distorted patch, respectively. Assuming that the residual errors follow the multi-variate Gaussian distributions, the self-complexity of the reference and transformational complexity between the reference and distorted samples are computed using covariance matrices. Additionally, the prediction terms generated by SA-VAR are introduced as one auxiliary feature to promote the final quality prediction. The effectiveness of the proposed transformational complexity based distortion metric (TCDM) is evaluated through extensive experiments conducted on five public point cloud quality assessment databases. The results demonstrate that TCDM achieves state-of-the-art (SOTA) performance, and further analysis confirms its robustness in various scenarios.

A Wideband and Compact Millimeter-Wave Antenna Array Fed by Printed Ridge Gap Waveguide for 5G Applications

This paper proposes a printed ridge gap waveguide (PRGW)-based 2 × 2 millimeter-wave antenna array with broad bandwidth and compact aperture. A dual-resonance element consisting of a quarter circular and a quarter annular patch is initially designed. Subsequently, to reduce the cross polarization (XP) radiation and enhance the gain of the proposed antenna, a four-element array with symmetric arrangement is developed, leading to a decline in the normalized XP levels from -4.5 to -18.9 dB and from -4.4 to -18.7 dB in the xoz and yoz planes, respectively. In addition, a pair of parasitic shorted half-circular patches is loaded to improve the antenna gain in the highfrequency range. Finally, a PRGW feeding network is designed and applied to provide differential feeding for the planar array. To validate the reflection and radiation performance, an array prototype is fabricated, and experimental verification is conducted. Measurement results show that a -10-dB impedance bandwidth of 14.3% (26.0–30.0 GHz) is realized for the prototype, and an average realized gain of 9.8 dBi with low XP radiation is achieved by the proposed array.

Dataset of Registered Hematoxylin–Eosin and Ki67 Histopathological Image Pairs Complemented by a Registration Algorithm

In this work, we describe a dataset suitable for analyzing the extent to which hematoxylin–eosin (HE)-stained tissue contains information about the expression of Ki67 in immunohistochemistry staining. The dataset provides images of corresponding pairs of HE and Ki67 stainings and is complemented by algorithms for computing the Ki67 index. We introduce a dataset of high-resolution histological images of testicular seminoma tissue. The dataset comprises digitized histology slides from 77 conventional testicular seminoma patients, obtained via surgical resection. For each patient, two physically adjacent tissue sections are stained: one with hematoxylin and eosin, and one with Ki67 immunohistochemistry staining. This results in a total of 154 high-resolution images. The images are provided in PNG format, facilitating ease of use for image analysis compared to the original scanner output formats. Each image contains enough tissue to generate thousands of non-overlapping 224 × 224 pixel patches. This shows the potential to generate more than 50,000 pairs of patches, one with HE staining and a corresponding Ki67 patch that depicts a very similar part of the tissue. Finally, we present the results of applying a ResNet neural network for the classification of HE patches into categories according to their Ki67 label.

Identification of metacommunities in bioregions with historical habitat networks.

Although metacommunity theory provides many useful insights for conservation planning, the transfer of this knowledge to practice is hampered due to the difficulty of identifying metacommunities in bioregions. This study aims to identify the spatial extent of metacommunities at bioregional scales using current and historical habitat data, especially because contemporary biodiversity patterns may be a result of time-lagged responses to historical habitat configurations. Further, this estimation of the metacommunity spatial extent is based on both the habitat structure and the dispersal ability of the species. Focusing on dragonfly and damselfly (odonate) species in the eastern Swiss Plateau, the research uses wetland habitat information spanning over 110 years to create a time series of nine habitat networks between 1899 and 2010. From these networks, we identified the spatial extents of metacommunities based on the year of habitat information as well as on watershed boundaries. To identify the best metacommunity spatial extents, the study investigates whether patch pairs within a metacommunity exhibit greater similarity in species composition (i.e. lower beta-diversity) than patch pairs between metacommunities. For the different metacommunities, we further investigated correlations between gamma diversity and metacommunity size and compare them to theoretical expectations. In both analyses we found that augmenting spatial metacommunity identification with historical geographical proximity results in stronger associations with biodiversity patterns (beta and gamma diversity) than when using only current-day habitat or watershed information.

A Study on a Compact Double Layer Sub-GHz Reflectarray Design Suitable for Wireless Power Transfer

The paper presents a novel small-footprint varactor diode-based reconfigurable reflectarray (RRA) design and investigates its power reflection efficiency theoretically and experimentally in a real-life indoor environment. The surface is designed to operate at 865.5 MHz and is intended for simultaneous use with other wireless power transfer (WPT) efficiency-improving techniques that have been recently reported in the literature. To the best of the authors’ knowledge, no RRA intended to improve the performance of antenna-based WPT systems operating in the sub-GHz range has been designed and studied both theoretically and experimentally so far. The proposed RRA is a two-layer structure. The top layer contains electronically tunable phase shifters for the local phase control of an incoming electromagnetic wave, while the other one is fully covered by metal to reduce the phase shifter size and RRA’s backscattering. Each phase shifter is a pair of diode-loaded 8-shaped metallic patches. Extensive numerical studies are conducted to ascertain a suitable set of RRA unit cell parameters that ensure both adequate phase agility and reflection uniformity for a given varactor parameter. The RRA design parameter finding procedure followed in this paper comprises several steps. First, the phase and amplitude responses of a virtual infinite double periodic RRA are computed using full-wave solver Ansys HFSS. Once the design parameters are found for a given set of physical constraints, the phase curve of the corresponding finite array is retrieved to estimate the side lobe level due to the finiteness of the RRA aperture. Then, a diode reactance combination is found for several different RRA reflection angles, and the corresponding RRA radiation pattern is computed. The numerical results show that the side lobe level and the deviation of the peak reflected power angles from the desired ones are more sensitive to the reflection coefficient magnitude uniformity than to the phase agility. Furthermore, it is found that for scanning angles less than 50°, satisfactory reflection efficiency can be achieved by using the classical reactance profile synthesis approach employing the generalized geometrical optics (GGO) approximation, which is in accord with the findings of other studies. Additionally, for large reflection angles, an alternative synthesis approach relying on the Floquet mode amplitude optimization is utilized to verify the maximum achievable efficiency of the proposed RRA at large angles. A prototype consisting of 36 elements is fabricated and measured to verify the proposed reflectarray design experimentally. The initial diode voltage combination is found by applying the GGO-based phase profile synthesis method to the experimentally obtained phase curve. Then, the voltage combination is optimized in real time based on power measurement. Finally, the radiation pattern of the prototype is acquired using a pair of identical 4-director printed Yagi antennas with a gain of 9.17 dBi and compared with the simulated. The calculated results are consistent with the measured ones. However, some discrepancies attributed to the adverse effects of biasing lines are observed.

Dynamics of a 3D Piezo-Magneto-Elastic Energy Harvester with Axisymmetric Multi-Stability.

In this investigation, a three-dimensional (3D) axisymmetric potential well-based nonlinear piezoelectric energy harvester is proposed to increase the broadband frequency response under low-strength planar external excitation. Here, a two-dimensional (2D) planar bi-stable Duffing potential is generalized into three dimensions by utilizing axial symmetry. The resulting axisymmetric potential well has infinitely many stable equilibria and one unstable equilibria at the highest point of the potential barrier for this cantilevered oscillator. Dynamics of such a 3D piezoelectric harvester with axisymmetric multi-stability are studied under planar circular excitation motion. Bifurcations of average power harvested from the two pairs of piezoelectric patches are presented against the frequency variation. The results show the presence of several branches of large-amplitude cross-well type period-1 and subharmonic solutions. Subharmonics involved in such responses are verified from the Fourier spectra of the solutions. The identified subharmonic solutions perform interesting patterns of curvilinear oscillations, which do not cross the potential barrier through its highest point. These solutions can completely or partially avoid the climbing of the potential barrier, thereby requiring low input excitation energy for barrier crossing. The influence of excitation amplitude on the bifurcations of normalized power is also investigated. Through multiple solution branches of subharmonic solutions, producing comparable power to the period-1 branch, broadband frequency response characteristics of such a 3D axisymmetically multi-stable harvester are highlighted.

PostBP: A Python library to analyze outputs from wildfire growth models

Wildfire is an important natural disturbance agent in Canadian forests, but it has also caused significant economic damage nationwide. Spatial fire growth models have emerged as important tools for representing wildfire dynamics across diverse landscapes, enabling the mapping of key wildfire hazard metrics such as location-specific burn probabilities or likelihoods of fire ignition. While these summary metrics have gained popularity, they often fall short in capturing the directional spread of wildfires and their potential spread distances. The metrics depicting the directional spread of wildfire can be derived from raw outputs generated with fire growth models, such as the perimeters and ignition locations of individual fires, but extracting this information requires complex data processing. To address this data gap, we present PostBP, an open-source Python package designed for post-processing the raw outputs of fire growth models — the ignition locations and perimeters of individual fires simulated over multiple stochastic iterations — into a matrix of fire spread likelihoods between all pairs of forest patches in a landscape. The PostBP also generates several other summary outputs, such as the source-sink ratio and the fire spread rose diagram. We provide an overview of PostBP's capabilities and demonstrate its practical application to a forested landscape.•Wildfire growth models generate large amounts of outputs, which are hard to summarize for practical decision-making.•The PostBP package calculates the summary metrics characterizing the directional spread of wildfires.•The fire risk summaries generated with PostBP can support the assessments of wildfire risk and mitigation measures.

VINCENT: Cyber-threat detection through vision transformers and knowledge distillation

Vision Transformers (ViTs) denote a family of attention-based deep learning techniques that have recently achieved amazing results in various problems related to the field of computer vision. In this paper, we explore the use of ViTs in problems of cyber-threat detection related to malware and network intrusion detection. In particular, we propose VINCENT, that is a novel deep neural method, which resorts to a color imagery representation of cyber-data by encoding related cyber-data features into neighboring color pixels. ViTs are trained from cyber-data images as teacher models, to extract explainable imagery signatures of cyber-data classes. This knowledge is extracted by leveraging the self-attention mechanism to give paired attention values between pairs of imagery patches. The signature knowledge, extracted through the ViT teacher, is, finally, used to train a smaller neural student model according to the knowledge distillation theory. Experiments with various benchmark cybersecurity datasets assess the accuracy of the student model VINCENT also compared to that of several state-of-the-art methods. In addition, it shows that VINCENT can obtain insights from explanations recovered through the self-attention mechanism of the ViT teacher.

A compact double‐sided four‐port printed antenna with pattern and polarization diversity for super‐wideband applications

SummaryA compact, double‐sided four‐port printed antenna with pattern and polarization diversity performance is proposed for super‐wideband (SWB) applications. The proposed printed antenna consists of microstrip line‐fed corner truncated four identical square patches. One pair of patches is placed strategically on the top side of the substrate, and another pair is on the bottom side with rotational symmetry to realize a compact four‐port antenna with higher port isolation (>25 dB) throughout the operating band along with pattern diversity. A half‐circular slit is etched in the partial ground plane of all four antennas to achieve SWB (1–19.81 GHz) response. A meander‐shaped decoupling structure is incorporated on both sides of the substrate to enhance the port isolation. T‐shaped slots on the patch and corner truncations are conceived to achieve circular polarization (CP) across 5.15–5.83 GHz. The beauty of this structure lies in the easy switching between LHCP and RHCP by just controlling the port excitation. Moreover, the strategic placement of antenna elements produces bidirectional patterns with proper choice of port excitation. A good agreement is observed between the measured and simulation results.

The Gauge Picture of Quantum Dynamics

Although local Hamiltonians exhibit local time dynamics, this locality is not explicit in the Schrödinger picture in the sense that the wavefunction amplitudes do not obey a local equation of motion. We show that geometric locality can be achieved explicitly in the equations of motion by "gauging" the global unitary invariance of quantum mechanics into a local gauge invariance. That is, expectation values ⟨ψ|A|ψ⟩ are invariant under a global unitary transformation acting on the wavefunction |ψ⟩→U|ψ⟩ and operators A→UAU†, and we show that it is possible to gauge this global invariance into a local gauge invariance. To do this, we replace the wavefunction with a collection of local wavefunctions |ψJ⟩, one for each patch of space J. The collection of spatial patches is chosen to cover the space; e.g. we could choose the patches to be single qubits or nearest-neighbor sites on a lattice. Local wavefunctions associated with neighboring pairs of spatial patches I and J are related to each other by dynamical unitary transformations UIJ. The local wavefunctions are local in the sense that their dynamics are local. That is, the equations of motion for the local wavefunctions |ψJ⟩ and connections UIJ are explicitly local in space and only depend on nearby Hamiltonian terms. (The local wavefunctions are many-body wavefunctions and have the same Hilbert space dimension as the usual wavefunction.) We call this picture of quantum dynamics the gauge picture since it exhibits a local gauge invariance. The local dynamics of a single spatial patch is related to the interaction picture, where the interaction Hamiltonian consists of only nearby Hamiltonian terms. We can also generalize the explicit locality to include locality in local charge and energy densities.

PatchCL-AE: Anomaly detection for medical images using patch-wise contrastive learning-based auto-encoder

Anomaly detection is an important yet challenging task in medical image analysis. Most anomaly detection methods are based on reconstruction, but the performance of reconstruction-based methods is limited due to over-reliance on pixel-level losses. To address the limitation, we propose a patch-wise contrastive learning-based auto-encoder for medical anomaly detection. The key contribution is the patch-wise contrastive learning loss that provides supervision on local semantics to enforce semantic consistency between corresponding input–output patches. Contrastive learning pulls corresponding patch pairs closer while pushing non-corresponding ones apart between input and output, enabling the model to learn local normal features better and improve discriminability on anomalous regions. Additionally, we design an anomaly score based on local semantic discrepancies to pinpoint abnormalities by comparing feature difference rather than pixel variations. Extensive experiments on three public datasets (i.e., brain MRI, retinal OCT, and chest X-ray) achieve state-of-the-art performance, with our method achieving over 99% AUC on retinal and brain images. Both the contrastive patch-wise supervision and patch-discrepancy score provide targeted advancements to overcome the weaknesses in existing approaches.

UFCC: A Unified Forensic Approach to Locating Tampered Areas in Still Images and Detecting Deepfake Videos by Evaluating Content Consistency

Image inpainting and Deepfake techniques have the potential to drastically alter the meaning of visual content, posing a serious threat to the integrity of both images and videos. Addressing this challenge requires the development of effective methods to verify the authenticity of investigated visual data. This research introduces UFCC (Unified Forensic Scheme by Content Consistency), a novel forensic approach based on deep learning. UFCC can identify tampered areas in images and detect Deepfake videos by examining content consistency, assuming that manipulations can create dissimilarity between tampered and intact portions of visual data. The term “Unified” signifies that the same methodology is applicable to both still images and videos. Recognizing the challenge of collecting a diverse dataset for supervised learning due to various tampering methods, we overcome this limitation by incorporating information from original or unaltered content in the training process rather than relying solely on tampered data. A neural network for feature extraction is trained to classify imagery patches, and a Siamese network measures the similarity between pairs of patches. For still images, tampered areas are identified as patches that deviate from the majority of the investigated image. In the case of Deepfake video detection, the proposed scheme involves locating facial regions and determining authenticity by comparing facial region similarity across consecutive frames. Extensive testing is conducted on publicly available image forensic datasets and Deepfake datasets with various manipulation operations. The experimental results highlight the superior accuracy and stability of the UFCC scheme compared to existing methods.

A Compact Low-profile 5G Millimeter-wave Circularly Polarized Antenna Based on LTCC

In this paper, a circularly polarized millimeter-wave L-shaped dipole antenna based on low temperature cofired ceramic (LTCC) technology is proposed, which realizes compact size and low-profile performance. The designed antenna consists of radiation patches and the grounded coplanar waveguide-substrate integrated waveguide (GCPW-SIW) feeding structure, which connects each other by two via holes. The radiation patches include a pair of L-shaped patches and four parasitic patches. The simulated results show that the proposed antenna operates from 26.5 to 29.5 GHz for |S11|<−10 dB and AR<3 dB with a peak gain of 6.7 dBic. The antenna element size is only 0.58λ0×0.58λ0×0.056λ0, where λ0 is free-space wavelength at the center frequency of 28 GHz. A sample of the antenna is fabricated and measured to verify the proposed design, which has a good agreement with the simulated ones, indicating that the antenna has potential applications for the fifth generation (5G) mm-Wave n257 (26.5 - 29.5 GHz) frequency band communications and satellite communication systems.

High isolation MIMO antenna designed with tightly coupled microstrip patch pairs

This paper proposes an array antenna with high isolation for MIMO application. Two tightly coupled microstrip patch pairs are proposed as array element with two rectangular stubs connected. By adjusting the position and width of the stubs, good resonant frequency matching can be achieved, and isolation can be effectively improved across the operating frequency band. Furthermore, 1 × 2 and 1 × 4 arrays with only 0.04 wavelength space between the edges are constructed for MIMO application. By adding an open-loop resonators (OLR) and slots in the metal reflector, the isolation between adjacent ports can be improved to 26 dB. In order to verify the feasibility of the proposed array antenna, the array antenna has been simulated, processed and tested, respectively. The simulation and measurement agree well and show the isolation more than 25 dB for all ports. The proposed structure has great potential in practical MIMO applications.

A broadband circularly polarized antenna for millimeter-wave applications

Abstract This paper presents a broadband circularly polarized (CP) antenna array for millimeter-wave applications, and the antenna array has the advantages of wide impedance bandwidth (IBW), novel CP design, and low profile. The antenna unit consists of a two-layer substrate and two pairs of magnetoelectric dipoles. Stepped microstrip lines coupled by rectangular slits form a feeder network for easy integration. The axial ratio bandwidth (ARBW) is extended because a pair of parasitic patches is loaded and an elliptical perturbation is added. The simulation results show that the antenna has an ARBW of 18.6% (26.4–31.9 GHz) and an IBW of 45.5% (20.6–32.7 GHz), with a gain greater than 7.11 dBic in the IBW. To improve the gain of the antenna, a 2 × 2 antenna array is designed, fabricated, and measured. The measured results show that the array has an ARBW of 16.6% (26.42–31.21 GHz), an IBW of 41.6% (22.28–33.97 GHz), a peak gain of 13.89 dBic in the IBW, the cross-polarization levels in the xoz-plane and yoz-plane are above 20 dB, and a radiation efficiency greater than 89%.

Dynamic Spatial Focus for Efficient Compressed Video Action Recognition

Recent years have witnessed a growing interest in compressed video action recognition due to the rapid growth of online videos. It remarkably reduces the storage by replacing raw videos with sparsely sampled RGB frames and other compressed motion cues (motion vectors and residuals). However, existing compressed video action recognition methods face two main issues: First, the inefficiency caused by the usage of coarse-level information under full resolution, and second, the disturbing due to the noisy dynamics in motion vectors. To address the two issues, this paper proposes a dynamic spatial focus method for efficient compressed video action recognition (CoViFocus). Specifically, we first use a light-weighted two-stream architecture to localize the task-relevant patches for both the RGB frames and motion vectors. Then the selected patch pair will be processed by a high-capacity two-stream deep model for the final prediction. Such a patch selection strategy crops out the irrelevant motion noise in motion vectors, as well as reduces the spatial redundancy of the inputs, leading to the high efficiency of our method in the compressed domain. Moreover, we found that the motion vectors can help our method to address the possibly happened <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">static-issue</i> , which means that the focus patches get stuck at some regions related to static objects rather than target actions, which further improves our method. Extensive results on both the HMDB-51 and UCF-101 datasets demonstrate the effectiveness and efficiency of our method in compressed video action recognition tasks.

A deep learning approach to remove contrast from contrast-enhanced CT for proton dose calculation.

Non-Contrast Enhanced CT (NCECT) is normally required for proton dose calculation while Contrast Enhanced CT (CECT) is often scanned for tumor and organ delineation. Possible tissue motion between these two CTs raises dosimetry uncertainties, especially for moving tumors in the thorax and abdomen. Here we report a deep-learning approach to generate NCECT directly from CECT. This method could be useful to avoid the NCECT scan, reduce CT simulation time and imaging dose, and decrease the uncertainties caused by tissue motion between otherwise two different CT scans. A deep network was developed to convert CECT to NCECT. The network receives a 3D image from CECT images as input and generates a corresponding contrast-removed NCECT image patch. Abdominal CECT and NCECT image pairs of 20 patients were deformably registered and 8000 image patch pairs extracted from the registered image pairs were utilized to train and test the model. CTs of clinical proton patients and their treatment plans were employed to evaluate the dosimetric impact of using the generated NCECT for proton dose calculation. Our approach achieved a Cosine Similarity score of 0.988 and an MSE value of 0.002. A quantitative comparison of clinical proton dose plans computed on the CECT and the generated NCECT for five proton patients revealed significant dose differences at the distal of beam paths. V100% of PTV and GTV changed by 3.5% and 5.5%, respectively. The mean HU difference for all five patients between the generated and the scanned NCECTs was ∼4.72, whereas the difference between CECT and the scanned NCECT was ∼64.52, indicating a ∼93% reduction in mean HU difference. A deep learning approach was developed to generate NCECTs from CECTs. This approach could be useful for the proton dose calculation to reduce uncertainties caused by tissue motion between CECT and NCECT.

Taxonomic review of the Calotesversicolor complex (Agamidae, Sauria, Squamata) in China, with description of a new species and subspecies.

Caloteswangisp. nov., a new species of the agamid genus Calotes Cuvier, 1817, from southern China and northern Vietnam, is described. This species can be distinguished from all known congeners by a combination of morphological characteristics and genetic divergence in the mitochondrial tRNA, ND2, and CO1 genes. Molecular phylogenetic analysis revealed that the new species was formed as a monophyletic group and that considerable genetic divergence existed between its congeners (minimum p-distance, 4.6%). Caloteswangisp. nov. is distinguished by a combination of the following characteristics: average SVL < 90 mm for adult males; 10-14 dorsal eyelid scales; scales on side of neck and adjacent shoulder area pointing obliquely upward; keels on neck scales weakly to strongly developed; fold in front of the shoulder absent; pair of dark triangular patches extending from the front of the shoulder to the jaw angles; and orange coloration of the tongue. Caloteswangisp. nov. is similar to C.irawadi but differs in having scales between the nasal shield and the orbit and a fourth toe with a claw that can reach between the eyes and tympanum (even to the snout when hind the limbs are adpressed forward). Phylogenetic analyses revealed two well-supported subspecies, Lineages A and B in C.wangisp. nov., with mean uncorrected p-distances between them of 2%. We propose that Lineage A, which is mainly from the central and southern Wuzhi Mountains on Hainan Island, is a subspecies, C.w.hainanensisssp. nov. Lineage B mainly comprises individuals from other sites on the island plus the adjacent mainland, and is described as subspecies, C.w.wangissp. nov. A diagnostic key to all Calotes species of China is also provided.

TokenCut: Segmenting Objects in Images and Videos with Self-supervised Transformer and Normalized Cut.

In this paper, we describe a graph-based algorithm that uses the features obtained by a self-supervised transformer to detect and segment salient objects in images and videos. With this approach, the image patches that compose an image or video are organised into a fully connected graph, in which the edge between each pair of patches is labeled with a similarity score based on the features learned by the transformer. Detection and segmentation of salient objects can then be formulated as a graph-cut problem and solved using the classical Normalized Cut algorithm. Despite the simplicity of this approach, it achieves state-of-the-art results on several common image and video detection and segmentation tasks. For unsupervised object discovery, this approach outperforms the competing approaches by a margin of 6.1%, 5.7%, and 2.6% when tested with the VOC07, VOC12, and COCO20K datasets. For the unsupervised saliency detection task in images, this method improves the score for Intersection over Union (IoU) by 4.4%, 5.6% and 5.2%. When tested with the ECSSD, DUTS, and DUT-OMRON datasets. This method also achieves competitive results for unsupervised video object segmentation tasks with the DAVIS, SegTV2, and FBMS datasets. Our implementation is available at https://www.m-psi.fr/Papers/TokenCut2022/.