This paper presents a new bit-plane-wise unequal error protection algorithm for progressive bitstreams transmitted over lossy networks. The proposed protects compressed embedded bitstream generated by 3-D SPIHT assigning an amount of forward correction (FEC) to each bit-plane. reduces the side information needed send size code decoder limiting number quality levels bit-planes be sent while providing graceful degradation picture as packet losses increase. We also apply our transmission JPEG...

Multi-class and multi-scale object detection for autonomous driving is challenging because of the high variation in scales cluttered background complex street scenes. Context information high-resolution features are keys to achieve a good performance detection. However, context typically unevenly distributed, feature map also contains distractive low-level features. In this paper, we propose location-aware deformable convolution backward attention filtering improve performance. The extracts...

This paper presents a distributed video streaming framework using unbalanced multiple description coding (MDC) and unequal error protection. In the proposed framework, two senders simultaneously stream complementary descriptions to single receiver over different paths. To minimize overall distortion exploit benefits of multipath transport when characteristics each path are different, an MDC method for wavelet-based coders combined with TCP-friendly rate allocation algorithm is proposed. The...

Universal image segmentation aims to handle all tasks within a single model architecture and ideally requires only one training phase. To achieve task-conditioned joint training, task token needs be used in the multi-task condition for specific tasks. Existing approaches generate from text input (e.g., "the is panoptic"). However, such text-based inputs merely serve as labels fail capture inherent differences between tasks, potentially misleading model. In addition, discrepancy visual...

With the introduction of Microsoft Kinect into gaming industry and release Kinect-based application development kits, a whole new has evolved around with applications for gaming, gesture recognition controlling devices, 3-D communication, etc. coming to fore. The popularity sensor can be attributed its low cost real-time depth map generation capability. However, owing Kinect's lack sophistication, maps generated have lot artifacts like poorly object boundaries missing values misalignment...

Recent advancements in image segmentation have been notably driven by Vision Transformers. These transformer-based models offer one versatile network structure capable of handling a variety tasks. Despite their effectiveness, the pursuit enhanced capabilities often leads to more intricate architectures and greater computational demands. OneFormer has responded these challenges introducing query-text contrastive learning strategy active during training only. However, this approach not...

Generating an accurate and dense disparity image is one of the important requirements for many applications such as 3D video stereo vision-based advanced driver assistance systems (ADAS). Depth estimation process obtaining a depth map based on two or more reference images. Recently, several techniques that use semi-global optimization estimating maps have been suggested. Although robustness against illumination changes vital factor in like ADAS, matching (SGM) mutual information achieves...

Video object detection enhances the performance of still-image based by exploiting temporal context information from neighboring frames. Most state-of-the-art video detectors are non-causal and require lots preceding succeeding frames, which makes them impractical for real-time online where frames not available. In this paper, we propose a causal recurrent flow-based method detection. The proposed reads only current frame one memory buffer at each time step. Two types utilized. short-term is...

Transformer-based semantic segmentation methods have achieved excellent performance in recent years. Mask2Former is one of the well-known transformer-based which unifies common image into a universal model. However, it performs relatively poorly obtaining local features and segmenting small objects due to relying heavily on transformers. To this end, we propose simple yet effective architecture that introduces auxiliary branches during training capture dense encoder side. The obtained help...

This paper presents a TCP-friendly rate allocation algorithm for multiple description coding combined with path diversity. A 3-D SPIHT-based is used to generate two independent substreams and each substream protected an FEC-based unequal error protection algorithm. In the proposed video streaming framework, senders simultaneously stream complementary receiver over different paths. The coordinates distributed from single minimize overall distortion. Simulation results show that increases...

In this paper we introduce Slowee, a smart eating-speed guide system with light and vibration. Slowee aims to improve the user's eating habits by delivering right feedback in real time user while eating. We designed implemented our system, conducted pilot study investigate usability of obtain feedbacks from users. Although number participants is rather small (n=10), gave positive on potentials Slowee. expect that device can help maintaining appropriate speed chewing numbers for patients...

Quality‐of‐service‐guaranteed video communication in wireless sensor networks (WVSNs) is extremely challenging because of the unique constraints WVSN (e.g. limited resources sensors and high error rates networks) characteristics traffic huge data rate tight delay bounds). Distributed coding (DVC) has emerged as a new paradigm for applications with resource‐limited encoders. However, current DVC architectures still have several technical limitations that prevent their widespread use...

One of the major challenges in video object detection is drastic scale changes objects due to camera motion.In this paper, we propose a two-path Convolutional Long Short-Term Memory (convLSTM) pyramid network designed extract and convey multi-scale temporal contextual information order handle efficiently.The proposed convLSTM consists stack multi-input modules.It updated top-down bottom-up pathways so that for small-to-large large-to-small exploited.The module uses two input feature maps...

In this paper, an adaptive motion vector smoothing scheme based on weighted median filtering is proposed in order to eliminate the outliers more effectively for improving quality of side information frame-based distributed video coding. We use a simple outlier reliability measure each block compensated interpolated frame and apply only blocks with unreliable vectors. Simulation results show that algorithm improves significantly while maintaining low complexity at encoder

Compressed sensing (CS) is the theory and practice of sub-Nyquist sampling sparse signals interest. Exact reconstruction may then be possible with much fewer than Nyquist-required number data. In this paper, we consider a multiview video system in which multiple cameras at different locations perform independent CS to simultaneously capture views scene. At decoder, propose disparity- motion-compensated total variation minimization algorithm jointly reconstruct sequence. The experimental...

In this paper we introduce a feedback system for the prevention of forward head posture in sedentary work environments. Our aims to promote user maintain good by delivering real-time when poses postures particularly sitting at desk. We designed and developed visual using 3D camera simple pop-up notification window. The results pilot study with 14 participants show that window made positive influence on preventing his or her posture.