Low-level Details Research Articles

Semantic-Guided Information Alignment Network for Fine-Grained Image Recognition

Existing <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">fine-grained image recognition</i> works have attempted to dig into low-level details for emphasizing subtle discrepancies among sub-categories. However, a potential limitation of these methods is that they integrate the low-level details and high-level semantics directly, and neglect their content complementarity and spatial corresponding correlation. To handle this limitation, we propose an end-to-end Semantic-guided Information Alignment Network (SIA-Net) to dynamically pick out the low-level details under the guidance of accurate semantics to make selected details spatially corresponding to high-level semantics and complementary in content. Technically, SIA-Net consists of an Accurate Semantic Calibration (ASC) module for providing accurate semantics and a Discriminative Feature Alignment (DFA) module for aggregating low-level details and high-level semantics using accurate semantics generated by ASC. ASC learns the pixel-level feature shifting caused by convolutional operations, which is utilized for replacing the incorrectly highlighted semantics by shifting discriminative semantics or background features. After obtaining the accurate semantic features, DFA digs into the complementary details and simultaneously makes the selected details spatially corresponding via applying the guidance of accurate semantics to obtain the reassembly features. Finally, the reassembly features, which serve as discriminative cues, are used for more accurate discriminative region localization. Extensive experiments verify that our proposed method yields the best performance under the same settings with the most competitive approaches on CUB-birds, Stanford-Cars, and FGVC Aircraft datasets.

HL-Pow: Learning-Assisted Pre-RTL Power Modeling and Optimization for FPGA HLS

High-level synthesis (HLS) enables designers to customize hardware designs without the need for delving into low-level hardware details. However, it is still challenging to establish the correlation between power consumption and hardware designs at an early design stage such as HLS. To overcome this problem, we introduce HL-Pow, a pre-register-transfer-level (pre-RTL) power modeling and optimization framework for FPGA HLS with the aid of up-to-date artificial intelligence techniques, which features high accuracy, speed and generalization ability. HL-Pow is comprised of a power modeling framework and a design space exploration (DSE) engine. The power modeling framework encompasses (1) a fully customized and light-weight feature construction flow to effectively identify and capture features that exert a major influence on power consumption; and (2) a modeling flow that can build an accurate, fast and transferable pre-RTL power estimator. With HL-Pow, the power evaluation process for hardware designs with FPGA HLS can be significantly expedited by circumventing the invocation of the time-consuming logic synthesis, physical design and gate-level simulation steps. Furthermore, we describe a novel a priori knowledge-guided DSE algorithm which can combined with our power modeling approach to jointly achieve the design optimization for latency and power consumption with high efficiency and high quality. Experimental results demonstrate that HL-Pow produces accurate power prediction that is only 4.82% away from onboard power measurement, while offering a speedup of 24–190× (84× on avg.). In addition, HL-Pow shows high generalization ability across applications with different characteristics and from various domains. Finally, the proposed DSE algorithm can reach a close approximation of the real Pareto frontier while only requiring traversing a small subset of design points in a broad design space.

EPT-Net: Edge Perception Transformer for 3D Medical Image Segmentation.

The convolutional neural network has achieved remarkable results in most medical image seg- mentation applications. However, the intrinsic locality of convolution operation has limitations in modeling the long-range dependency. Although the Transformer designed for sequence-to-sequence global prediction was born to solve this problem, it may lead to limited positioning capability due to insufficient low-level detail features. Moreover, low-level features have rich fine-grained information, which greatly impacts edge segmentation decisions of different organs. However, a simple CNN module is difficult to capture the edge information in fine-grained features, and the computational power and memory consumed in processing high-resolution 3D features are costly. This paper proposes an encoder-decoder network that effectively combines edge perception and Transformer structure to segment medical images accurately, called EPT-Net. Under this framework, this paper proposes a Dual Position Transformer to enhance the 3D spatial positioning ability effectively. In addition, as low-level features contain detailed information, we conduct an Edge Weight Guidance module to extract edge information by minimizing the edge information function without adding network parameters. Furthermore, we verified the effectiveness of the proposed method on three datasets, including SegTHOR 2019, Multi-Atlas Labeling Beyond the Cranial Vault and the re-labeled KiTS19 dataset called KiTS19-M by us. The experimental results show that EPT-Net has significantly improved compared with the state-of-the-art medical image segmentation method.

Improved UNet with Attention for Medical Image Segmentation.

Medical image segmentation is crucial for medical image processing and the development of computer-aided diagnostics. In recent years, deep Convolutional Neural Networks (CNNs) have been widely adopted for medical image segmentation and have achieved significant success. UNet, which is based on CNNs, is the mainstream method used for medical image segmentation. However, its performance suffers owing to its inability to capture long-range dependencies. Transformers were initially designed for Natural Language Processing (NLP), and sequence-to-sequence applications have demonstrated the ability to capture long-range dependencies. However, their abilities to acquire local information are limited. Hybrid architectures of CNNs and Transformer, such as TransUNet, have been proposed to benefit from Transformer's long-range dependencies and CNNs' low-level details. Nevertheless, automatic medical image segmentation remains a challenging task due to factors such as blurred boundaries, the low-contrast tissue environment, and in the context of ultrasound, issues like speckle noise and attenuation. In this paper, we propose a new model that combines the strengths of both CNNs and Transformer, with network architectural improvements designed to enrich the feature representation captured by the skip connections and the decoder. To this end, we devised a new attention module called Three-Level Attention (TLA). This module is composed of an Attention Gate (AG), channel attention, and spatial normalization mechanism. The AG preserves structural information, whereas channel attention helps to model the interdependencies between channels. Spatial normalization employs the spatial coefficient of the Transformer to improve spatial attention akin to TransNorm. To further improve the skip connection and reduce the semantic gap, skip connections between the encoder and decoder were redesigned in a manner similar to that of the UNet++ dense connection. Moreover, deep supervision using a side-output channel was introduced, analogous to BASNet, which was originally used for saliency predictions. Two datasets from different modalities, a CT scan dataset and an ultrasound dataset, were used to evaluate the proposed UNet architecture. The experimental results showed that our model consistently improved the prediction performance of the UNet across different datasets.

AMCFNet: Asymmetric multiscale and crossmodal fusion network for RGB-D semantic segmentation in indoor service robots

Red-green-blue and depth (RGB-D) semantic segmentation is essential for indoor service robots to achieve accurate artificial intelligence. Various RGB-D indoor semantic segmentation methods have been proposed since the widespread adoption of depth maps. These methods have focused mainly on integrating the multiscale and crossmodal features extracted from RGB images and depth maps in the encoder and used unified strategies to recover the local details at the decoder progressively. However, these methods emphasized crossmodal fusion at the encoder, neglecting the distinguishability between RGB and depth features during decoding, thereby undermining the segmentation performance. To adequately exploit the features, we propose an efficient encoder-decoder architecture called asymmetric multiscale and crossmodal fusion network (AMCFNet) endowed with a differential feature integration strategy. Unlike existing methods, we use simple crossmodal fusion at the encoder and design an elaborate decoder to improve the semantic segmentation performance. Specifically, considering high- and low-level features, we propose a semantic aggregation module (SAM) to process the multiscale and crossmodal features in the last three network layers to aggregate high-level semantic information through a cascaded pyramid structure. Moreover, we design a spatial detail supplement module using low-level spatial details from depth maps to adaptively fuse these details and the information obtained from the SAM. Extensive experiments are conducted to demonstrate that the proposed AMCFNet outperforms state-of-the-art approaches.

Co-learning-assisted progressive dense fusion network for cardiovascular disease detection using ECG and PCG signals

Electrocardiograms (ECGs) and phonocardiograms (PCGs) are two modalities to provide complementary diagnostic information for improving the early detection accuracy of cardiovascular diseases (CVDs). Existing multi-modality methods mainly used the early or late feature fusion strategy, which did not simultaneously utilize the complementary information contained in low-level detail features and high-level semantic features of different modalities. Meanwhile, they were specially designed for the multi-modality scenario with both ECGs and PCGs, without considering the missing-modality scenarios with only ECGs or PCGs in clinical practice. To address these challenges, we developed a Co-learning-assisted Progressive Dense fusion network (CPDNet) for end-to-end CVD detection, with a three-branch interweaving architecture consisting of ECG and PCG modality-specific encoders and a progressive dense fusion encoder, which could be used for both multi-modality and missing-modality scenarios. Specifically, we designed a novel progressive dense fusion strategy, which not only progressively fused multi-level complementary information of different modalities from low-level details to high-level semantics, but also employed the dense fusion during feature fusion at each level to further enrich available multi-modality information through mutual guidance of features at different levels. Meanwhile, the strategy integrated cross-modality region-aware and multi-scale feature optimization modules to fully evaluate the contributions of different modalities and signal regions and enhance the feature extraction ability of the network for multi-scale target regions. Moreover, we designed a novel co-learning strategy to guide the learning process of the CPDNet by combining intra-modality and joint losses, which made each encoder well-trained. This strategy could not only assist our fusion strategy by making modality-specific encoders provide sufficiently discriminative features for the fusion encoder, but also enable the CPDNet to robustly handle missing-modality scenarios by independently using the corresponding modality-specific encoder. Experimental results on public and private datasets demonstrated that our method not only outperformed state-of-the-art multi-modality methods by at least 5.05% for average accuracy in the multi-modality scenario, but also achieved better performance than single-modality models in the missing-modality scenarios.

Improving defocus blur detection via adaptive supervision prior-tokens

The Defocus Blur Detection (DBD) technique is devised to accurately identify regions of blurriness within images. The prediction difficulty of defocused pixels is closely associated with their spatial location. Owing to the cluttered background, pixels near the edges are more prone to erroneous predictions. To address the issue of uneven pixel distribution at the edges of defocused regions, we deliberately decouple the original labels into Prior-Tokaens: Edge Transition Detail Region (EDR) and Structure Body Region (SBR). Subsequently, we propose a novel adaptive multi-supervised network comprising a feature extraction module, a feature fusion network (FFN), and a Multi-scale Channel Attention Module (MCAM). This method harnesses complementary features between SBR and EDR, furnishing a tailored feature learning strategy that outperforms traditional single-supervised techniques. Furthermore, considering that features generated with varying receptive fields contain information at different levels, we introduce MCAM to identify feature pixels at different scales, enhancing semantic relevance. Moreover, for images with complex scenes, an adaptive learning scheme is developed to selectively fuse low-level detail features and high-level semantic information, thereby enhancing the model's generalization capability. The proposed approach outperforms state-of-the-art techniques on various evaluation metrics, as demonstrated through qualitative and quantitative analyses of popular public datasets.

Multi-Scale Flame Situation Detection Based on Pixel-Level Segmentation of Visual Images

The accurate analysis of multi-scale flame development plays a crucial role in improving firefighting decisions and facilitating smart city establishment. However, flames’ non-rigid nature and blurred edges present challenges in achieving accurate segmentation. Consequently, little attention is paid to extracting further flame situation information through fire segmentation. To address this issue, we propose Flame-SeaFormer, a multi-scale flame situation detection model based on the pixel-level segmentation of visual images. Flame-SeaFormer comprises three key steps. Firstly, in the context branch, squeeze-enhanced axial attention (SEA attention) is applied to squeeze fire feature maps, capturing dependencies among flame pixels while reducing the computational complexity. Secondly, the fusion block in the spatial branch integrates high-level semantic information from the contextual branch with low-level spatial details, ensuring a global representation of flame features. Lastly, the light segmentation head conducts pixel-level segmentation on the flame features. Based on the flame segmentation results, static flame parameters (flame height, width, and area) and dynamic flame parameters (change rates of flame height, width, and area) are gained, thereby enabling the real-time perception of flame evolution behavior. Experimental results on two datasets demonstrate that Flame-SeaFormer achieves the best trade-off between segmentation accuracy and speed, surpassing existing fire segmentation methods. Flame-SeaFormer enables precise flame state acquisition and evolution exploration, supporting intelligent fire protection systems in urban environments.

Attention-based dual-path feature fusion network for automatic skin lesion segmentation

Automatic segmentation of skin lesions is a critical step in Computer Aided Diagnosis (CAD) of melanoma. However, due to the blurring of the lesion boundary, uneven color distribution, and low image contrast, resulting in poor segmentation result. Aiming at the problem of difficult segmentation of skin lesions, this paper proposes an Attention-based Dual-path Feature Fusion Network (ADFFNet) for automatic skin lesion segmentation. Firstly, in the spatial path, a Boundary Refinement (BR) module is designed for the output of low-level features to filter out irrelevant background information and retain more boundary details of the lesion area. Secondly, in the context path, a Multi-scale Feature Selection (MFS) module is constructed for high-level feature output to capture multi-scale context information and use the attention mechanism to filter out redundant semantic information. Finally, we design a Dual-path Feature Fusion (DFF) module, which uses high-level global attention information to guide the step-by-step fusion of high-level semantic features and low-level detail features, which is beneficial to restore image detail information and further improve the pixel-level segmentation accuracy of skin lesion. In the experiment, the ISIC 2018 and PH2 datasets are employed to evaluate the effectiveness of the proposed method. It achieves a performance of 0.890/ 0.925 and 0.933 /0.954 on the F1-score and SE index, respectively. Comparative analysis with state-of-the-art segmentation methods reveals that the ADFFNet algorithm exhibits superior segmentation performance.

B-ultrasound guided venipuncture vascular recognition system based on deep learning

Venipuncture is one of the common operations used by doctors in the outpatient blood drawing room. The success rate of puncture is not only related to the pain degree of the patient, but also affects the test results. In this study, we propose an ultrasound guided venipuncture vascular recognition system based on deep learning. First, kmeans++ clustering is performed for the vascular regions in the different B-mode ultrasound images to facilitate estimation in subsequent work. Second, a lightweight vascular ultrasound network (UV-YOLOv7) is designed, specifically, based on YOLOv7-tiny, a multi-scale feature fusion module (MFFM) is designed to better fuse the high-level semantic features and low-level detail features, and the speed and accuracy of model detection are enhanced by lightweighting the model structure and replacing the EIoU loss function. Finally, a Dynamic Neighborhood-Density Based Spatial Clustering of Applications with Noise (DN-DBSCAN) algorithm is proposed, which can cluster a series of local vascular regions using the localization results and confidence properties of the network output to remove the misdetected regions. In the experiment, We selected 303 artifact-free and 264 heavily artifacted vascular ultrasound images for offline expansion and trained on the experimental platform, The results show that the proposed method performed best with an mAP of 86.2% and an inference time of 0.6 ms. At the end of the experiment, more robust vascular localization results were obtained by DN-DBSCAN clustering.

Neural representations for multi-context visuomotor adaptation and the impact of common representation on multi-task performance: a multivariate decoding approach.

The human brain's remarkable motor adaptability stems from the formation of context representations and the use of a common context representation (e.g., an invariant task structure across task contexts) derived from structural learning. However, direct evaluation of context representations and structural learning in sensorimotor tasks remains limited. This study aimed to rigorously distinguish neural representations of visual, movement, and context levels crucial for multi-context visuomotor adaptation and investigate the association between representation commonality across task contexts and adaptation performance using multivariate decoding analysis with fMRI data. Here, we focused on three distinct task contexts, two of which share a rotation structure (i.e., visuomotor rotation contexts with -90° and +90° rotations, in which the mouse cursor's movement was rotated 90 degrees counterclockwise and clockwise relative to the hand-movement direction, respectively) and the remaining one does not (i.e., mirror-reversal context where the horizontal movement of the computer mouse was inverted). This study found that visual representations (i.e., visual direction) were decoded in the occipital area, while movement representations (i.e., hand-movement direction) were decoded across various visuomotor-related regions. These findings are consistent with prior research and the widely recognized roles of those areas. Task-context representations (i.e., either -90° rotation, +90° rotation, or mirror-reversal) were also distinguishable in various brain regions. Notably, these regions largely overlapped with those encoding visual and movement representations. This overlap suggests a potential intricate dependency of encoding visual and movement directions on the context information. Moreover, we discovered that higher task performance is associated with task-context representation commonality, as evidenced by negative correlations between task performance and task-context-decoding accuracy in various brain regions, potentially supporting structural learning. Importantly, despite limited similarities between tasks (e.g., rotation and mirror-reversal contexts), such association was still observed, suggesting an efficient mechanism in the brain that extracts commonalities from different task contexts (such as visuomotor rotations or mirror-reversal) at multiple structural levels, from high-level abstractions to lower-level details. In summary, while illuminating the intricate interplay between visuomotor processing and context information, our study highlights the efficiency of learning mechanisms, thereby paving the way for future exploration of the brain's versatile motor ability.

A whole-task brain model of associative recognition that accounts for human behavior and neuroimaging data.

Brain models typically focus either on low-level biological detail or on qualitative behavioral effects. In contrast, we present a biologically-plausible spiking-neuron model of associative learning and recognition that accounts for both human behavior and low-level brain activity across the whole task. Based on cognitive theories and insights from machine-learning analyses of M/EEG data, the model proceeds through five processing stages: stimulus encoding, familiarity judgement, associative retrieval, decision making, and motor response. The results matched human response times and source-localized MEG data in occipital, temporal, prefrontal, and precentral brain regions; as well as a classic fMRI effect in prefrontal cortex. This required two main conceptual advances: a basal-ganglia-thalamus action-selection system that relies on brief thalamic pulses to change the functional connectivity of the cortex, and a new unsupervised learning rule that causes very strong pattern separation in the hippocampus. The resulting model shows how low-level brain activity can result in goal-directed cognitive behavior in humans.

Entering the Metaverse from the JVM: The State of the Art, Challenges, and Research Areas of JVM-Based Web 3.0 Tools and Libraries

Web 3.0 is the basis on which the proposed metaverse, a seamless virtual world enabled by computers and interconnected devices, hopes to interact with its users, but beyond the high-level project overview of what Web 3.0 applications try to achieve, the implementation is still down to low-level coding details. This article aims to analyze the low-level implementations of key components of Web 3.0 using a variety of frameworks and tools as well as several JVM-based languages. This paper breaks down the low-level implementation of smart contracts and semantic web principles using three frameworks, Corda and Ethereum for smart contracts and Jeda for semantic web, using both Scala and Java as implementing languages all while highlighting differences and similarities between the frameworks used.

Scene sketch semantic segmentation with hierarchical Transformer

Convolutional Neural Networks (CNNs) have always been the dominant method for scene sketch semantic segmentation, but their performance seems to have plateaued due to the limitation of local receptive fields. To address this problem, we propose SketchSeger, a hierarchical Transformer-based model for scene sketch semantic segmentation. Accurate scene sketch segmentation relies on both high-level semantics and low-level details. To obtain better segmentation performance, we designed an MLP-based feature fusion module for the model decoder to merge feature maps captured at different scales efficiently. Compared to CNN-based models, SketchSeger exhibits a stronger ability in contextual modeling and can obtain global receptive fields even in its shallow layers. Besides the model architecture, the absence of large-scale pre-training datasets also presents a significant challenge for advancing scene sketch semantic segmentation. To promote further research, we propose a novel hand-drawn style scene sketch synthesis method and use it to synthesize a dataset containing 300,000 annotated scene sketches. We conduct extensive experiments and visual analysis to validate the efficacy of our proposed SketchSeger model and dataset synthesis approach. The results show that SketchSeger significantly outperforms state-of-the-art models on three benchmark datasets (SketchyScene, SKY-Scene, and TUB-Scene) with similar parameter scales. Codes and datasets are available at https://github.com/jayangcs/SketchSeger.

Influence of equipment, weather and personnel experience on the digitalization and damage detection of concrete bridges

AbstractTraditional visual inspections of existing infrastructure have limitations such as safety risks, limited accuracy for subsurface defects, and subjective assessments. Modern inspection techniques like terrestrial laser scanning, closerange photogrammetry (CRP), and infrared scanning offer faster and more accurate data collection, allowing for more informative and precise descriptions of objects, while minimizing traffic disturbances often associated with visual inspections. In this paper, CPR, a non‐destructive survey technique for reconstructing 3D shapes from 2D images, is used for the inspection of five bridges in Northern Sweden. The surveys were performed under different weather conditions, using different data acquisition equipment and personnel with different levels of experience. The acquired images were used to create 3D models of the bridges and to evaluate the evolution of damages over time. Results showed that although better models were created when using equipment with higher specifications, cheap gear such as GoPro or cellphones can also be used for the cases in which a lower level of detail is required. Furthermore, initial experience of the personnel does not seem to have high influence on the results.

Multilevel Feature Fusion for End-to-End Blind Image Quality Assessment

In this paper, a framework based on two feature extraction networks and a multilevel feature fusion (MFF) network is proposed. Multilevel degradation features can be obtained through this method, and combined with the human visual perception system, the local and global feature information contained in these features can be captured, which is conducive to the prediction of distorted images. First, a restored image approximating a reference image is generated by a restorative generative adversarial network (GAN). Furthermore, the multilevel degradation features of distorted images and the restored image features are extracted by EfficientNet. Second, the features extracted by EfficientNet are input into the MFF network and are fully expressed by the top-down, bottom-up and third edge joining methods. Moreover, the features provide more low-level details and high-level semantic features for the prediction of image quality scores. In addition, after the MFF stage, the framework calculates the score of each branch feature and obtains the average quality score. Experimental results show that our method achieves greatly improved prediction accuracy and performance on five standard databases.

MUSE: Visual Analysis of Musical Semantic Sequence.

Visualization has the capacity of converting auditory perceptions of music into visual perceptions, which consequently opens the door to music visualization (e.g., exploring group style transitions and analyzing performance details). Current research either focuses on low-level analysis without constructing and comparing music group characteristics, or concentrates on high-level group analysis without analyzing and exploring detailed information. To fill this gap, integrating the high-level group analysis and low-level details exploration of music, we design a musical semantic sequence visualization analytics prototype system (MUSE) that mainly combines a distribution view and a semantic detail view, assisting analysts in obtaining the group characteristics and detailed interpretation. In the MUSE, we decompose the music into note sequences for modeling and abstracting music into three progressively fine-grained pieces of information (i.e., genres, instruments and notes). The distribution view integrates a new density contour, which considers sequence distance and semantic similarity, and helps analysts quickly identify the distribution features of the music group. The semantic detail view displays the music note sequences and combines the window moving to avoid visual clutter while ensuring the presentation of complete semantic details. To prove the usefulness and effectiveness of MUSE, we perform two case studies based on real-world music MIDI data. In addition, we conduct a quantitative user study and an expert evaluation.

Dual-path segmentation network for automatic fabric defect detection

Fabric defect detection plays a crucial role in the production process of the textile industry. Vision-based inspection methods have emerged as an inevitable trend due to their lower labor costs and high detection efficiency. As the accuracy requirements for fabric defect detection, methods must not only identify and locate defects accurately but also describe the morphological features of the defects. This poses a challenge for the algorithm’s design, as it must consider both the semantic and texture information of the fabric. In this paper, we propose an end-to-end dual-path segmentation network called DPNet for fabric defect detection, which can extract and fuse both semantic and texture information to achieve high accuracy. The proposed framework consists of two paths: the semantic path, which has narrow but deep layers to obtain high-dimensional features, and the texture path, which has wide and dense layers to extract low-level details. To enhance the interaction between semantic and texture features, a crossed attention fusion module has been developed. Evaluations show that the proposed method outperforms other methods on different datasets in terms of mIoU, with results of 75.84% for Ngan and 70.69% for AITEX. In addition, we developed an inspection platform and tested the proposed method online. We found that it can achieve online detection at a speed of 40 m/min, making it well-suited for practical production environments.

Multi-Scale Semantic and Detail Extraction Network for Lightweight Person Re-Identification

Exploring multi-level information to obtain fine-grained features is the key factor to improve the performance of person re-identification (Re-ID). However, existing models for person Re-ID only focus on learning the high-level semantic information while neglecting the low-level detail information. To alleviate this issue, we propose a lightweight person Re-ID method termed Multi-Scale Semantic and Detail Extraction Network (MSDENet) to obtain robustness and discriminative feature representation for the Re-ID task. Specifically, we design a Series Channel-Spatial Attention (SCSA) and embed it into the lightweight backbone network to focus on the key parts of pedestrian images. Meanwhile, we propose a Multi-Scale Semantic and Detail Extraction (MSDE) method to extract multi-scale features of semantic information and detail information, which can effectively capture the feature diversity of pedestrian images. Furthermore, we design a Feature Enhancement Fusion (FEF), which enhances and fuses the fine-grained features of semantic extraction and detail extraction branches to better obtain the discriminative feature representation. Extensive experiments conducted on popular datasets Market1501, MSMT17, and CUHK03 demonstrate that the proposed MSDENet has competitive performance compared with the state-of-the-art methods.

Cross Pyramid Transformer makes U-net stronger in medical image segmentation

Accurate auto-medical image segmentation, which is essential in disease diagnosis and treatment planning, has been significantly prospered by recent advances in Convolution Neural Network (CNN) and Transformer. However, pure CNN-based or pure Transformer-base architectures exhibit limitations for segmentation tasks. For example, CNN fails to capture long-range relations due to fix receptive fields, and Transformer ignores pixel-level spatial details of features. To address these challenges, we propose a novel parallel hybrid architecture for medical image segmentation, which is named CPT-Unet (Cross Pyramid Transformer U-shape Network). CPT-Unet may be the first attempt to exploit both the advantages of Pyramid Vision Transformer (PVT) and CNN to the full by integrating them into the standard U-shape network to improve the segmentation performance and inference time. Specifically, we design a parallel dual branch encoder and decoder in CPT-Unet that consists of CNN and PVT. The input image is fed into the encoder of these two branches simultaneously to extract low-level spatial details and global contexts in a much shallower way. We design a novel fusion strategy to adequately utilize the multi-scale features extracted from CNN and PVT. We also add PVT in the decoder to better restore the segmentation map. Experiments on two public segmentation datasets demonstrate the improved performance of the proposed CPT-Unet over the comparison methods. The source code is available at https://github.com/ShengYue007/CPT-Unet.git.