Search

Data augmentation is widely recognized as an effective means of bolstering model robustness. However, when applied to monocular 3D object detection, non-geometric image augmentation neglects the critical link between the image and physical space, resulting in the semantic collapse of the extended scene. To address this issue, we propose two geometric-level data augmentation operators named Geometric-Copy-Paste (Geo-CP) and Geometric-Crop-Shrink (Geo-CS). Both operators introduce geometric consistency based on the principle of perspective projection, complementing the options available for data augmentation in monocular 3D. Specifically, Geo-CP replicates local patches by reordering object depths to mitigate perspective occlusion conflicts, and Geo-CS re-crops local patches for simultaneous scaling of distance and scale to unify appearance and annotation. These operations ameliorate the problem of class imbalance in the monocular paradigm by increasing the quantity and distribution of geometrically consistent samples. Experiments demonstrate that our geometric-level augmentation operators effectively improve robustness and performance in the KITTI and Waymo monocular 3D detection benchmarks. [ABSTRACT FROM AUTHOR]

Background: Skin incision scars are cosmetically displeasing; the effects of current treatments are limited, and new methods to reduce scar formation need to be found. Objective: We sought to determine whether immediate postoperative injection of stromal vascular fraction gel (SVF-gel) could reduce scar formation at skin incision sites. Methods: A prospective, randomized, double-blind, self-controlled trial was conducted in patients who underwent breast reduction. SVF-gel was intradermally injected into the surgical incision on one randomly selected side, with the other side receiving saline as a control. At the 6-month follow-up, the incision scars were evaluated using the Vancouver scar scale (VSS) and visual analog scale (VAS). Antera 3D camera was used for objective evaluation. Results: The VSS score and VAS score were significantly different between the SVF-gel-treated side (3.80 ± 1.37, 3.37±1.25) and the control side (5.25 ± 1.18, 4.94 ± 1.28). Moreover, the SVF-gel-treated side showed statistically significant improvements in scar appearance, based on evidences from Antera 3D camera. Limitations: This was a single-center, single-race, and single-gender study. Furthermore, the results were available only for the 6-month interim follow-up period. Conclusion: Postoperative immediate SVF-gel injection in surgical incisions can reduce scar formation, and exert a preventive effect on scars. Level of Evidence I: Evidence obtained from at least one properly designed randomized controlled trial. This journal requires that authors assign a level of evidence to each article. For a full description of these Evidence-Based Medicine ratings, please refer to the Table of Contents or the online Instructions to Authors www.springer.com/00266. [ABSTRACT FROM AUTHOR]

Artificial intelligence (AI) systems surpass certain human intelligence abilities in a statistical sense as a whole, but are not yet the true realization of these human intelligence abilities and behaviors. There are differences, and even contradictions, between the cognition and behavior of AI systems and humans. With the goal of achieving general AI, this study contains a review of the role of cognitive science in inspiring the development of the three mainstream academic branches of AI based on the three-layer framework proposed by David Marr, and the limitations of the current development of AI are explored and analyzed. The differences and inconsistencies between the cognition mechanisms of the human brain and the computation mechanisms of AI systems are analyzed. They are found to be the cause of the differences and contradictions between the cognition and behavior of AI systems and humans. Additionally, eight important research directions and their scientific issues that need to focus on brain-inspired AI research are proposed: highly imitated bionic information processing, a large-scale deep learning model that balances structure and function, multi-granularity joint problem solving bidirectionally driven by data and knowledge, AI models that simulate specific brain structures, a collaborative processing mechanism with the physical separation of perceptual processing and interpretive analysis, embodied intelligence that integrates the brain cognitive mechanism and AI computation mechanisms, intelligence simulation from individual intelligence to group intelligence (social intelligence), and AI-assisted brain cognitive intelligence. [ABSTRACT FROM AUTHOR]

Background: Autologous adipose tissue often experiences ischemia and hypoxia after transplantation, leading to low retention rates and unstable operative impacts due to necrotic absorption. Platelet-rich plasma (PRP) can enhance fat regeneration and increase the fat retention rate after transplantation. However, the quick release of growth factors (GFs) in PRP decreases therapeutic efficiency. This study aimed to achieve a slow release of PRP to promote fat retention. Methods: We prepared a dual-network hydrogel (DN gel) based on FDA-approved PRP and sodium alginate (SA) through a simple "one-step" activation process. In vivo study, adipose tissue with saline (control group), SA gel (SA gel group), PRP gel (PRP gel group), and DN gel (DN gel group) was injected subcutaneously into the dorsum of nude mice. At 4 and 12 weeks after injection, tissues were assessed for volume and weight. Hematoxylin and eosin staining (HE) and immunofluorescence staining were performed for histological assessment. Results: DN gel exhibits long-lasting growth factor effects, surpassing conventional clinical PRP gel regarding vascularization potential. In fat transplantation experiments, DN gel demonstrated improved vascularization of transplanted fat and increased retention rates, showing promise for clinical applications. Conclusions: DN gel-assisted lipofilling can significantly improve the retention rate and quality of transplanted fat. DN gel-assisted lipofilling, which is considered convenient, is a promising technique to improve neovascularization and fat survival. No Level Assigned: This journal requires that authors assign a level of evidence to each submission to which Evidence-Based Medicine rankings are applicable. This excludes Review Articles, Book Reviews, and manuscripts that concern Basic Science, Animal Studies, Cadaver Studies, and Experimental Studies. For a full description of these Evidence-Based Medicine ratings, please refer to the Table of Contents or the online Instructions to Authors www.springer.com/00266 [ABSTRACT FROM AUTHOR]

Weakly supervised semantic segmentation (WSSS) aims to reduce the cost of collecting dense pixel-level annotations for segmentation models by adopting weak labels to train. Although WSSS methods have achieved great success, recent approaches mainly concern the image-level label-based WSSS, which is limited to object-centric datasets instead of more challenging practical datasets that contain many co-occurrent classes. In comparison, point-level labels could provide some spatial information to address the class co-occurrent confusion problem. Meanwhile, it only requires an additional click when recognizing the targets, which is of negligible annotation overhead. Thus, we choose to study utilizing point labels for the general-purpose WSSS. The main difficulty of utilizing point-level labels is bridging the gap between the sparse point-level labels and the dense pixel-level predictions. To alleviate this problem, we propose a superpixel augmented pseudo-mask generation strategy and a class-aware contrastive learning approach, which manages to recover reliable dense constraints and apply them both to the segmentation models' final prediction and the intermediate features. Diagnostic experiments on the challenging Pascal VOC, Cityscapes, and the ADE20k datasets demonstrate that our approach can efficiently and effectively compensate for the sparse point-level labels and achieve cutting-edge performance on the point-based segmentation problems. [ABSTRACT FROM AUTHOR]

Tremendous efforts have been made on image segmentation but the mask quality is still not satisfactory. The boundaries of predicted masks are usually imprecise due to the low spatial resolution of feature maps and the imbalance problem caused by the extremely low proportion of boundary pixels. To address these issues, we propose a conceptually simple yet effective post-processing refinement framework, termed BPR, to improve the boundary quality of the prediction of any image segmentation model. Following the idea of looking closer to segment boundaries better, we extract and refine a series of small boundary patches along the predicted boundaries. The refinement is accomplished by a boundary patch refinement network at the higher resolution. The trained BPR model can be easily transferred to refine the results of other models as well. Extensive experiments show that the proposed BPR framework yields significant improvements on the semantic, instance, and panoptic segmentation tasks over a variety of baselines on the Cityscapes dataset. [ABSTRACT FROM AUTHOR]

Scale-sensitive object detection remains a challenging task, where most of the existing methods could not learn it explicitly and are not robust. Besides, they are less efficient during training or slow during inference, which is not friendly to real-time applications. In this paper, we propose a scale-transferrable architecture for practical object detection based on the analysis of the connection between dilation rate and effective receptive field. Our method firstly predicts a global continuous scale, which is shared by all positions, for each convolution filter of each network stage. Secondly, we average the spatial features and distill the scale from channels to effectively learn the scale. Thirdly, for fast-deployment, we propose a scale decomposition method that transfers the robust fractional scale into the combination of fixed integral scales for each convolution filter, which exploits the dilated convolution. Moreover, to overcome the shortcomings of our method for large-scale object detection, we modify the Feature Pyramid Network structure. Finally, we illustrate the orthogonality role of our method for sampling strategy. We demonstrate the effectiveness of our method on one-stage and two-stage algorithms under different configurations and compare them with different dilated convolution blocks. For practical applications, the training strategy of our method is simple and efficient, avoiding complex data sampling or optimization strategy. During inference, we reduce the latency of the proposed method by using the hardware accelerator TensorRT without extra operation. On the COCO test-dev, our model achieves 41.7% mAP on one-stage detector and 42.5% mAP on two-stage detector based on ResNet-101, and outperforms baselines by 3.2% and 3.1% mAP, respectively. [ABSTRACT FROM AUTHOR]

The recording process of observation is influenced by multiple factors, such as viewpoint, illumination, and state of the object-of-interest etc.Thus, the image observation of the same object may vary a lot under different conditions. This leads to severe intra-class variance which greatly challenges the discrimination ability of the vision model. However, the current prevailing softmax loss for visual recognition only pursues perfect inter-class separation in the feature space. Without considering the intra-class compactness, the learned model easily collapses when it encounters the instances that deviate a lot from their class centroid. To resist the intra-class variance, we start by organizing the input instances as a graph. From this viewpoint, we find that the normalized cut on the graph is a favorable surrogate metric of the intra-class variance within the training batch. Inspired by the equivalence between the normalized cut and random walk, we propose a feature aggregation scheme using transition probabilities as guidance. By imposing supervision on the aggregated features, we can constrain the transition probabilities to form a graph partition consistent with the given labels. Thus, the normalized cut as well as intra-class variance can be well suppressed. To validate the effectiveness of this idea, we instantiate it in spatial, temporal, and spatial-temporal scenarios. Experimental results on corresponding benchmarks demonstrate that the proposed feature aggregation leads to significant improvement in performance. Our method is on par with, or even better than current state-of-the-arts in both tasks. [ABSTRACT FROM AUTHOR]

The mechanism of hard surfaces worn by soft polymers is not clearly understood. In this paper, a new hypothesis has been proposed, it holds that the stress acting on the hard surface under certain working conditions is the main reason for wear of the hard surface by a soft polymer. The hypothesis was investigated by changing the contact form between tribo-pairs. For this, friction tests between six polymer spheres and smooth, rough, and inclined monocrystalline silicon surfaces were carried out. The results show that for the same tribo-pair, the silicon surface will not be worn in some contact forms, but in other contact forms it will be worn. We believe the wear of hard surface by a soft polymer is the result of the combined stress state action on the hard surface. [ABSTRACT FROM AUTHOR]

Unsupervised domain adaptation has been a popular approach for cross-domain person re-identification (re-ID). There are two solutions based on this approach. One solution is to build a model for data transformation across two different domains. Thus, the data in source domain can be transferred to target domain where re-ID model can be trained by rich source domain data. The other solution is to use target domain data plus corresponding virtual labels to train a re-ID model. Constrains in both solutions are very clear. The first solution heavily relies on the quality of data transformation model. Moreover, the final re-ID model is trained by source domain data but lacks knowledge of the target domain. The second solution in fact mixes target domain data with virtual labels and source domain data with true annotation information. But such a simple mixture does not well consider the raw information gap between data of two domains. This gap can be largely contributed by the background differences between domains. In this paper, a Suppression of Background Shift Generative Adversarial Network (SBSGAN) is proposed to mitigate the gaps of data between two domains. In order to tackle the constraints in the first solution mentioned above, this paper proposes a Densely Associated 2-Stream (DA-2S) network with an update strategy to best learn discriminative ID features from generated data that consider both human body information and also certain useful ID-related cues in the environment. The built re-ID model is further updated using target domain data with corresponding virtual labels. Extensive evaluations on three large benchmark datasets show the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Domain adaptation (DA) for semantic segmentation aims to reduce the annotation burden for the dense pixel-level prediction task. It focuses on tackling the domain gap problem and manages to transfer knowledge learned from abundant source data to new target scenes. Although recent works have achieved rapid progress in this field, they still underperform fully supervised models with a large margin due to the absence of any available hints in the target domain. Considering that few-shot labels are cheap to obtain in practical applications, we attempt to leverage them to mitigate the performance gap between DA and fully supervised methods. The key to this problem is to leverage the few-shot labels to learn robust domain-invariant predictions effectively. To this end, we first design a data perturbation strategy to enhance the robustness of the representations. Furthermore, a transferable prototype module is proposed to bridge the domain gap based on the source data and few-shot targets. By means of these proposed methods, our approach can perform on par with the fully supervised models to some extent. We conduct extensive experiments to demonstrate the effectiveness of the proposed methods and report the state-of-the-art performance on two popular DA tasks, i.e., from GTA5 to Cityscapes and SYNTHIA to Cityscapes. [ABSTRACT FROM AUTHOR]

The deep inferior epigastric artery perforator (DIEP) flap has been widely used for approximately 30 years as an effective method for breast reconstruction, defect repair, and perineal reconstruction. However, few literatures have reported on thinned pedicle DIEP flaps for scrotal reconstruction. The present study presents a series of reconstruction of the scrotum using thinned pedicle DIEP flaps. Five patients with large scrotal skin defects were admitted to our hospital from October 2010 to December 2015. Among these patients, three patients had testes and spermatic cord exposure after accidental injury, while two patients had testes exposure after machine extrusion. All of these patients were repaired using thinned pedicle DIEP flaps. During the 12–24-month follow up, one patient experienced a bulky flap, and a subsequent thinned flap was successfully made. The other four flaps were successfully transplanted with good esthetic and functional results. All donor sites were primarily closed; there was no functional disturbance, such as dysuria and abdominal hernia, and the sensation of these flaps partially recovered after 6 months. The thinned pedicle DIEP flap is robust, reliable, and resilient, and offers good-quality skin cover and cushion to the testes. This flap can be considered a good choice for scrotal reconstruction. [ABSTRACT FROM AUTHOR]

With the development of the satellite spatial resolution, high-resolution and high-quality remote sensing images require high stability of the observation satellite platform. Therefore, real-time attitude information is essential to enhance the accuracy and the robustness of the attitude control system in observation satellites. Through the use of the parallax observation system, this paper designs a novel framework to estimate the on-orbit attitude, which includes a complete deconvolution process, a detailed camera motion model, and a support vector regression algorithm to detect the attitude and deal with the latency. The proposed framework is testified by realistically simulated data and remote sensing images from the JL-1 satellite. Simulation and experiment results demonstrate that the proposed framework can detect and predict the attitude in real-time while the error is lower than 1.125μrad. [ABSTRACT FROM AUTHOR]

This paper aims at solving online equality-constrained quadratic programming problem, which is widely encountered in science and engineering, e.g., computer vision and pattern recognition, digital signal processing, and robotics. Recurrent neural networks such as conventional GradientNet and ZhangNet are considered as powerful solvers for such a problem in light of its high computational efficiency and capability of circuit realisation. In this paper, an improved primal recurrent neural network and its electronic implementation are proposed and analysed. Compared to the existing recurrent networks, i.e. GradientNet and ZhangNet, our network can theoretically guarantee superior global exponential convergence. Robustness performance of our such neural model is also analysed under a large model implementation error, with the upper bound of stead-state solution error estimated. Simulation results demonstrate theoretical analysis on the proposed model, which also verify the effectiveness of the proposed model for online equality-constrained quadratic programming. [ABSTRACT FROM AUTHOR]

Traditional machine learning methods suffer from severe overfitting in EEG-based emotion reading. In this paper, we use hierarchical convolutional neural network (HCNN) to classify the positive, neutral, and negative emotion states. We organize differential entropy features from different channels as two-dimensional maps to train the HCNNs. This approach maintains information in the spatial topology of electrodes. We use stacked autoencoder (SAE), SVM, and KNN as competing methods. HCNN yields the highest accuracy, and SAE is slightly inferior. Both of them show absolute advantage over traditional shallow models including SVM and KNN. We confirm that the high-frequency wave bands Beta and Gamma are the most suitable bands for emotion reading. We visualize the hidden layers of HCNNs to investigate the feature transformation flow along the hierarchical structure. Benefiting from the strong representational learning capacity in the two-dimensional space, HCNN is efficient in emotion recognition especially on Beta and Gamma waves. [ABSTRACT FROM AUTHOR]

Plasticity in our brain offers us promising ability to learn and know the world. Although great successes have been achieved in many fields, few bio-inspired machine learning methods have mimicked this ability. Consequently, when meeting large-scale or time-varying data, these bio-inspired methods are infeasible, due to the reasons that they lack plasticity and need all training data loaded into memory. Furthermore, even the popular deep convolutional neural network (CNN) models have relatively fixed structures and cannot process time varying data well. Through incremental methodologies, this paper aims at exploring an end-to-end lifelong learning framework to achieve plasticities of both the feature and classifier constructions. The proposed model mainly comprises of three parts: Gabor filters followed by max pooling layer offering shift and scale tolerance to input samples, incremental unsupervised feature extraction, and incremental SVM trying to achieve plasticities of both the feature learning and classifier construction. Different from CNN, plasticity in our model has no back propogation (BP) process and does not need huge parameters. Our incremental models, including IncPCANet and IncKmeansNet, have achieved better results than PCANet and KmeansNet on minist and Caltech101 datasets respectively. Meanwhile, IncPCANet and IncKmeansNet show promising plasticity of feature extraction and classifier construction when the distribution of data changes. Lots of experiments have validated the performance of our model and verified a physiological hypothesis that plasticity exists in high level layer better than that in low level layer. [ABSTRACT FROM AUTHOR]

In this paper, we investigate some interesting identities on the Bernoulli, Euler, Hermite and generalized Gegenbauer polynomials arising from the orthogonality of generalized Gegenbauer polynomials in the generalized inner product [ABSTRACT FROM AUTHOR]

Structure information plays an important role in both object recognition and detection. This paper studies what visual structure is and addresses the problem of structure modeling and representation from two aspects: visual feature and topology model. Firstly, at feature level, we propose Local Structured Descriptor to capture the object's local structure effectively, and develop the descriptors from shape and texture information, respectively. Secondly, at topology level, we present a local structured model with a boosted feature selection and fusion scheme. All experiments are conducted on the challenging PASCAL Visual Object Classes (VOC) datasets from VOC2007 to VOC2010. Experimental results show that our method achieves very competitive performance. [ABSTRACT FROM AUTHOR]

Background: Current knowledge about the blood supply of the nipple-areola complex (NAC) has largely been derived from studies on cadavers or persons with breasts of normal size. The aim of this study was to identify and classify the NAC blood supply by computed tomographic angiography (CTA) examination in female volunteers with breast hypertrophy. Methods: CTA examination was performed on hypertrophic breasts of 23 female subjects. The main blood supplies were revealed through image data analyses. The dominant blood supply of the NAC and its vascular sources were identified and sorted. The detectable diameter threshold of blood vessels was set beyond 1.0 mm. Results: A total of 61 dominant blood vessels were identified. The source arteries were traced as the internal thoracic artery (ITA, 50.8%), lateral thoracic artery (LTA, 27.8%), thoracoacromial artery (TA, 14.8%), brachial artery (BA, 3.3%), and axillary artery (AA, 3.3%), and the corresponding reproducibility of these source vessels was 31, 37, 9, 4.3, and 4.3%, in all breasts. The intercostal artery (IA) was not identified as a dominant NAC supplying vessel in any CTA scan image. Twenty-six breasts had only one dominant artery, whereas 17 breasts showed multiple dominant blood supplies. Three breasts showed no dominant blood vessels of the NAC, with diameters greater than the detectable threshold of 1.0 mm, and 52.2% of the breasts demonstrated anatomically symmetrical patterns of blood supply for the NAC. Conclusions: The ITA, LTA, and TA are likely to be the main vessel sources, whereas the IA is unlikely to be the dominant vessel for NAC perfusion, on the basis of the studied breasts. An asymmetrical pattern of bilateral breast blood supply was demonstrated in a considerable portion of the females with breast hypertrophy in this study. Level of Evidence IV: This journal requires that authors assign a level of evidence to each article. For a full description of these Evidence-Based Medicine ratings, please refer to the Table of Contents or the online Instructions to Authors . [ABSTRACT FROM AUTHOR]

Video-based face recognition is a fundamental topic in image processing and video analysis, and presents various challenges and opportunities. In this paper, we introduce an incremental learning approach to video-based face recognition which efficiently exploits the spatiotemporal information in videos. Face image sequences are incrementally clustered based on their descriptors, and the representative face images of each cluster are picked out. The incremental algorithm of creating facial visual words is applied to construct a codebook using the descriptors of the representative face images. Continuously, with the quantization of the facial visual words, each descriptor extracted from patches is converted into codes, and codes from each region are pooled together into a histogram. The representation of the face image is generated by concatenating the histograms from all regions, which is employed to perform the categorization. In the online recognition, a similarity score matrix and a voting algorithm are employed to judge a face video's identity. Recognition is performed online while face video sequence is continuous and the proposed method gives nearly realtime feedback. The proposed method achieves a 100 % verification rate on the Honda/UCSD database and 82 % on the YouTube datebase. Experimental results demonstrate the effectiveness and flexibility of the proposed method. [ABSTRACT FROM AUTHOR]

We address the problem of on-line signature representation and verification. A novel strategy is proposed in this paper by combining texture based image analysis and spatio-temporal representation. Firstly, a correlation based method is proposed to describe spatio-temporal information between sampling points, which is then converted to traditional 2D intensity images. Secondly, abundant texture analysis methods are adopted to construct effective features for high accuracy verification. Furthermore, a template selection strategy based on intra-class variations is presented to further enhance the performance of signature verification. Extensive experiments are conducted on the SVC2004 database and experimental results demonstrate the inspiring performance of our proposed methods. [ABSTRACT FROM AUTHOR]

As malicious attacks greatly threaten the security and reliability of biometric systems, ensuring the authenticity of biometric data is becoming increasingly important. In this paper we propose a watermarking-based two-stage authentication framework to address this problem. During data collection, face features are embedded into a fingerprint image of the same individual as data credibility token and secondary authentication source. At the first stage of authentication, the credibility of input data is established by checking the validness of extracted patterns. Due to the specific characteristics of face watermarks, the face detection based classification strategies are introduced for reliable watermark verification instead of conventional correlation based watermark detection. If authentic, the face patterns can further serve as supplemental identity information to facilitate subsequential biometric authentication. In this framework, one critical issue is to guarantee the robustness and capacity of watermark while preserving the discriminating features of host fingerprints. Hence a wavelet quantization based watermarking approach is proposed to adaptively distribute watermark energy on significant DWT coefficients of fingerprint images. Experimental results which evaluate both watermarking and biometric authentication performance demonstrate the effectiveness of this work. [ABSTRACT FROM AUTHOR]

The performance of most gait recognition methods would drop down if the viewpoint of test data is different from the viewpoint of training data. In this paper, we present an idea of estimating the view angle of a test sample in advance so as to compare it with the corresponding training samples with the same or approximate viewpoint. In order to obtain reliable estimation results, the view-sensitive features should be extracted. We propose a novel and effective feature extraction method to characterize the silhouettes from different views. The discrimination power of this representation is also verified through experiments. Afterwards, the robust regression method is employed to estimate the viewpoint of gait. The view angles of test samples from BUAA-IRIP Gait Database are estimated with the regression models learned from CASIA Gait Database. Compared with the ground truth angles, such estimation is satisfactory with a small error level. Therefore, it can provide necessary help for gait application systems when the view angles of test data are uncertain. This point is verified experimentally through integrating the view angle estimation into a gait based gender classification system. [ABSTRACT FROM AUTHOR]