Accelerator-Aware Fast Spatial Feature Network for Real-Time Semantic Segmentation

Recent advances in convolutional neural networks have led to considerably accurate semantic segmentation tasks. However, previous works use a pre-trained backbone borrowed from image classification tasks with considerable computational complexity, resulting in large latencies. Many methods have been proposed to reduce the latency of the segmentation network without loss of accuracy. These methods have reported varying experimental results with different computing devices, acceleration techniques, and input image sizes. Although most studies claim that their results are state-of-the-art, it is necessary to reconsider whether they are being compared under the same conditions. We propose a performance evaluation method of real-time semantic segmentation models to compare the performance under the same conditions fairly. In addition, we carry out an empirical study to evaluate the performance of recent real-time semantic segmentation networks and make a comparative analysis between them. We train the segmentation models using the same input data and data augmentation method. Then, the performance of the segmentation methods is analyzed regarding accuracy and speed. In contrast to most studies that exclude the time required for the pre-processing and post-processing steps, we measured the actual processing time needed to perform semantic segmentation with a real dataset. Further, we measured the processing speed and power consumption of the segmentation models in embedded devices in which real-time segmentation is applied, unlike previous studies that measured performance on a PC. Experimental results showed that the real-time semantic segmentation methods could not run in real-time on embedded devices when considering the pre-processing and post-processing steps. By comprehensively considering the inference speed, energy consumption, and processing time of semantic segmentation models, the experimental results show that FasterSeg-S is suited for embedded devices.

Panoptic Segmentation aims to provide an understanding of background (stuff) and instances of objects (things) at a pixel level. It combines the separate tasks of semantic segmentation (pixel level classification) and instance segmentation to build a single unified scene understanding task. Typically, panoptic segmentation is derived by combining semantic and instance segmentation tasks that are learned separately or jointly (multi-task networks). In general, instance segmentation networks are built by adding a foreground mask estimation layer on top of object detectors or using instance clustering methods that assign a pixel to an instance center. In this work, we present a fully convolution neural network that learns instance segmentation from semantic segmentation and instance contours (boundaries of things). Instance contours along with semantic segmentation yield a boundary aware semantic segmentation of things. Connected component labeling on these results produces instance segmentation. We merge semantic and instance segmentation results to output panoptic segmentation. We evaluate our proposed method on the CityScapes dataset to demonstrate qualitative and quantitative performances along with several ablation studies. Our overview video can be accessed from https://youtu.be/wBtcxRhG3e0.

One of the bold issues in autonomous driving is considered semantic image segmentation, which must be done with high accuracy and speed. Semantic segmentation is used to understand an image at the pixel level. In this regard, various architectures based on deep neural networks have been proposed for semantic segmentation of autonomous driving image datasets. In this paper, we proposed a novel combination method in which dividing the image into its constituent regions with the help of classical segmentation brings about achieving beneficial information that improves the DeepLab v3+ network results. The proposed method with the two backbones, Xception and MobileNetV2, obtains the mIoU of 81.73% and 76.31% on the Cityscapes dataset, respectively, which shows promising results compared to the model without post-processing.

<div class="section abstract"><div class="htmlview paragraph">Traditional Simultaneous Localization and Mapping (SLAM) methods often assume static environments. This limitation can lead to inaccurate localization or even the loss of tracking in dynamic scenes. To address this issue, we propose a novel SLAM approach specifically designed for dynamic environments. Our method integrates the real-time image semantic segmentation network BisenetV2 with inter-frame and continuous multi-frame motion feature detection. Firstly, semantic segmentation is applied to render the semantic mask, which is then used by the inter-frame motion detection module to identify potential motion features. Subsequently, these suspected motion features are evaluated by a likelihood probability model across consecutive frames. Finally, points with a high probability of motion are monitored in real-time by the Luenberger observer, which filters out motion features and re-adds static ones. Our experiments demonstrate that semantic segmentation can meet real-time requirements in various scenarios. Evaluations of the TUM-RGBD dataset and simulation experiment show that our proposed method significantly improves SLAM accuracy and robustness in dynamic scenes.</div></div>

Vision Transformers (ViTs) have emerged as state-of-the-art models for various vision tasks recently. However, their heavy computation costs remain daunting for resource-limited devices. To address this, researchers have dedicated themselves to compressing redundant information in ViTs for acceleration. However, existing approaches generally sparsely drop redundant image tokens by token pruning or brutally remove channels by channel pruning, leading to a sub-optimal balance between model performance and inference speed. Moreover, they struggle when transferring compressed models to downstream vision tasks that require the spatial structure of images, such as semantic segmentation. To tackle these issues, we propose CAIT, a joint compression method for ViTs that achieves a harmonious blend of high accuracy, fast inference speed, and favorable transferability to downstream tasks. Specifically, we introduce an asymmetric token merging (ATME) strategy to effectively integrate neighboring tokens. It can successfully compress redundant token information while preserving the spatial structure of images. On top of it, we further design a consistent dynamic channel pruning (CDCP) strategy to dynamically prune unimportant channels in ViTs. Thanks to CDCP, insignificant channels in multi-head self-attention modules of ViTs can be pruned uniformly, significantly enhancing the model compression. Extensive experiments on multiple benchmark datasets show that our proposed method can achieve state-of-the-art performance across various ViTs.

In this study, we propose a segmentation model based on convolutional neural networks (CNNs) to address image segmentation challenges in computer vision. Prior to designing the model, the activation function and other modules of the convolutional neural network were optimized to meet specific requirements. The segmentation task was transformed into binary classification problem to simplify network calculations and improve efficiency. Additionally, the model utilized a mask map obtained from the semantic segmentation model to aid in instance segmentation. Class activation technology was introduced to extract feature mapping maps. The corresponding thermal maps were obtained to achieve target instance segmentation. To further validate the effectiveness of the segmentation model, simulation experiments were conducted on semantic segmentation and instance segmentation respectively. The results show that the accuracy of the basic semantic segmentation model reached 87.58%, while the average accuracy of the entire class of the optimized instance segmentation model reached 97.9%. Therefore, the research and design of image segmentation models demonstrate high accuracy and good robustness.

Recently, semantic segmentation based on deep neural network (DNN) has attracted attention as it exhibits high accuracy, and many studies have been conducted on this. However, DNN-based segmentation studies focused mainly on improving accuracy, thus greatly increasing the computational demand and memory footprint of the segmentation network. For this reason, the segmentation network requires a lot of hardware resources and power consumption, and it is difficult to be applied to an environment where they are limited, such as an embedded system. In this paper, we propose a binarized encoder-decoder network ( BEDN ) and a binarized deconvolution engine ( BiDE ) accelerating the network to realize low-power, real-time semantic segmentation. BiDE implements a binarized segmentation network with custom hardware, greatly reducing the hardware resource usage and greatly increasing the throughput of network implementation. The deconvolution used for upsampling in a segmentation network includes zero padding. In order to enable deconvolution in a binarized segmentation network that cannot express zero, we introduce zero-aware binarized deconvolution which skips padded zero activations and zero-aware batch normalization embedded binary activation considering zero-skipped convolution. The BEDN , which is a binarized segmentation network proposed to be accelerated on BiDE , has acceptable accuracy while greatly reducing the computational and memory demands of the segmentation network through full-binarization and simple structure. BEDN has a network size of 0.21 MB, and its maximum memory usage is 1.38 MB. BiDE was implemented on Xilinx ZU7EV field-programmable gate array (FPGA) to operate at 187.5 MHz. BiDE accelerated the proposed BEDN within CamVid11 images of $480\times {360}$ size at 25.89 frames per second (FPS) achieving a performance of 1.682 Tera operations per second (TOPS) and 824 Giga operations per second per watt (GOPS/W).

Unsupervised domain adaptive (UDA) semantic segmentation aims to transfer a semantic segmentation model from a labeled source domain to an unlabeled target domain. This paper focuses on UDA semantic segmentation in the underground railway scene, where no existing UDA works have been discussed. Due to the severe domain gap between source and target domains in the underground railway scene, existing methods in this field targeted at urban traffic scene have limitations to apply in the underground railway scene. This paper proposes a novel multi-level (Pixel level and Image sample level) pseudo-label correction method that can correct pseudo-labels in pixel and image sample levels. Our two stages pseudo-label correction method can significantly boost the quality of pseudo-labels and eventually boost the performance of UDA semantic segmentation in the underground railway scene. Experiments illustrate that the performance of our proposed method outperforms other existing methods in the underground railway scene.

With the growing demands of applications on online devices, the speed-accuracy trade-off is critical in the semantic segmentation system. Recently, the bilateral segmentation network has shown promising capacity to achieve the balance between favorable accuracy and fast speed, and has become the mainstream backbone in real-time semantic segmentation. Segmentation of target objects relies on high-level semantics, whereas it requires detailed low-level features to model specific local patterns for accurate location. However, the lightweight backbone of bilateral architecture limits the extraction of semantic context and spatial details. And the late fusion of the bilateral streams incurs the insufficient aggregation of semantic context and spatial details. In this article, we propose a densely aggregated bilateral network (DAB-Net) for real-time semantic segmentation. In the context path, a patchwise context enhancement (PCE) module is proposed to efficiently capture the local semantic contextual information from spatialwise and channelwise, respectively. Meanwhile, a context-guided spatial path (CGSP) is designed to exploit more spatial information by encoding finer details from the raw image and the transition from the context path. Finally, with multiple interactions between bilateral branches, the intertwined outputs from bilateral streams are combined in a unified decoder for a final interaction to further enhance the feature representation, which generates the final segmentation prediction. Experimental results on three public benchmarks demonstrate that our proposed method achieves higher accuracy with a limited decay in speed, which performs favorably against state-of-the-art real-time approaches and runs at 31.1 frames/s (FPS) on the high resolution of [Formula: see text] . The source code is released at https://github.com/isyangshu/DABNet.

As the techniques of autonomous driving become increasingly valued and universal, real-time semantic segmentation has become very popular and challenging in the field of deep learning and computer vision in recent years. However, in order to apply the deep learning model to edge devices accompanying sensors on vehicles, we need to design a structure that has the best trade-off between accuracy and inference time. In previous works, several methods sacrificed accuracy to obtain a faster inference time, while others aimed to find the best accuracy under the condition of real time. Nevertheless, the accuracies of previous real-time semantic segmentation methods still have a large gap compared to general semantic segmentation methods. As a result, we propose a network architecture based on a dual encoder and a self-attention mechanism. Compared with preceding works, we achieved a 78.6% mIoU with a speed of 39.4 FPS with a 1024 × 2048 resolution on a Cityscapes test submission.

Automatic extraction of cerebral vessels and cranial nerves has important clinical value in the treatment of trigeminal neuralgia (TGN) and hemifacial spasm (HFS). However, because of the great similarity between different cerebral vessels and between different cranial nerves, it is challenging to segment cerebral vessels and cranial nerves in real time on the basis of true-color microvascular decompression (MVD) images. In this paper, we propose a lightweight, fast semantic segmentation Microvascular Decompression Network (MVDNet) for MVD scenarios which achieves a good trade-off between segmentation accuracy and speed. Specifically, we designed a Light Asymmetric Bottleneck (LAB) module in the encoder to encode context features. A Feature Fusion Module (FFM) was introduced into the decoder to effectively combine high-level semantic features and underlying spatial details. The proposed network has no pretrained model, fewer parameters, and a fast inference speed. Specifically, MVDNet achieved 76.59% mIoU on the MVD test set, has 0.72 M parameters, and has a 137 FPS speed using a single GTX 2080Ti card.

Real-time semantic segmentation of images requires both rich contextual and accurate spatial information. However, Multiple downsampling in deep convolutional neural networks often lead to loss of such information, resulting in reduced segmentation accuracy. To address the above problems, we propose SPCONet, a lightweight real-time semantic segmentation network that integrates spatial and contextual features. The network incorporates three key modules: (1) a Spatial Feature Aggregation Module (SFAM) that captures fine spatial details from shallow layers using spatially separable convolutions with multiple kernel sizes; (2) a Contextual Information Retrieval Module (CIRM) that extracts semantic context from deeper layers using dynamic convolution; (3) an Attention Fusion Module (AFM) that combines spatial and contextual features via local and global attention mechanisms. Quantitative experiments show that SPCONet achieves 77.5% and 75.3% mIoU at 74 FPS and 82 FPS on the Cityscapes and CamVid datasets, respectively. These results suggest that SPCONet provides an effective balance between segmentation accuracy and real-time inference capability.

We present a novel, real-time, semantic segmentation network in which the encoder both encodes and generates the parameters (weights) of the decoder. Furthermore, to allow maximal adaptivity, the weights at each decoder block vary spatially. For this purpose, we design a new type of hypernetwork, composed of a nested U-Net for drawing higher level context features, a multi-headed weight generating module which generates the weights of each block in the decoder immediately before they are consumed, for efficient memory utilization, and a primary network that is composed of novel dynamic patch-wise convolutions. Despite the usage of less-conventional blocks, our architecture obtains real-time performance. In terms of the runtime vs. accuracy trade-off, we surpass state of the art (SotA) results on popular semantic segmentation benchmarks: PASCAL VOC 2012 (val. set) and real-time semantic segmentation on Cityscapes, and CamVid. The code is available: https://nirkin.com/hyperseg.

Research on image classification sparked the latest deep-learning boom. Many downstream tasks, including semantic segmentation, benefit from it. The state-of-the-art semantic segmentation models are all based on deep learning, and they sometimes make some semantic mistakes. In a semantic segmentation dataset with a small number of categories, images are often collected from a single scene, and there is a close semantic connection between any two categories. However, in the semantic segmentation dataset collected from multiple scenes, two categories may be irrelevant. The probability of objects in one category appearing next to objects in other categories is different, which is the basis of the paper. Semantic segmentation methods need to solve two problems of positioning and classification. This paper is dedicated to correcting those clearly wrong classifications that are contrary to reality. Specifically, we first calculate the relevancy between different class pairs. Then, based on this knowledge, we infer the category of a connected component according to the relationships of this connected component with its surrounding connected components and correct the obviously wrong classifications made by a deep learning semantic segmentation model. Several well-performing deep learning models are experimented on two challenging public datasets in the field of semantic image segmentation. Our proposed method improves the performance of UPerNet, OCRNet and SETR from 40.7%, 43% and 48.64% to 42.07%, 44.09% and 49.09% mean IoU on the ADE20K validation set, and the performance of PSPNet, DeepLabV3 and OCRNet from 37.26%, 37.3% and 39.5% to 38.93%, 38.95% and 40.63% mean IoU on the COCO-Stuff dataset, which shows the effectiveness of the method.

Real-time semantic segmentation via sequential knowledge distillation