Semantic segmentation is a fundamental task in indoor scene understanding. Most previous supervised approaches rely on densely annotated image data sets. Due to the limited amount of images with segmentation labels, t...
ISBN:
(数字)9781728123455
ISBN:
(纸本)9781728123462
Semantic segmentation is a fundamental task in indoor scene understanding. Most previous supervised approaches rely on densely annotated image data sets. Due to the limited amount of images with segmentation labels, the performance of existing networks is greatly limited. In this paper, we exploit temporal correlation in video frames to improve the performance and robustness of segmentation networks. Two effective learning strategies are proposed to propagate the information from a few labeled frames to their immediate neighbor frames. First, we scale up training dataset for supervised semantic segmentation networks by generating pseudo ground-truth for neighboring frames from a labeled frame using filtered homography transformation. Furthermore, we introduce a self-supervised loss function to ensure temporal consistency between the segmentation results of adjacent frames. The experimental results demonstrate that our proposed method outperforms state-of-the-art techniques for semantic segmentation on NYU-Depth V2 dataset.
Image compression had been extensively studied for reducing coding rate yet producing acceptable visual quality. However, there are many application scenarios where the compressed images are used for automatic recogni...
详细信息
ISBN:
(纸本)9781479973408
Image compression had been extensively studied for reducing coding rate yet producing acceptable visual quality. However, there are many application scenarios where the compressed images are used for automatic recognition rather than human viewing, thus the visual quality is no longer critical for compression. SIFT features have demonstrated their utility in many recognition scenarios and SIFT-preserving compression is developed recently. In this paper, we firstly study the SIFT-preserving compression of license plate images for recognition accuracy rather than visual quality. According to extracted SIFT features, each image is divided into SIFT coding-units and non-SIFT coding-units. Each coding-unit is assigned with a different quality parameter when using JPEG for compression. We compare our proposed scheme with the standard JPEG that uses a unified quality parameter. Experimental results with manually tuned parameters show that on average 14% bit-rate can be saved by our scheme, without any loss of recognition accuracy.
We have witnessed the rapid development of learned image compression (LIC). The latest LIC models have outperformed almost all traditional image compression standards in terms of rate-distortion (RD) performance. Howe...
详细信息
ISBN:
(纸本)9781728173221
We have witnessed the rapid development of learned image compression (LIC). The latest LIC models have outperformed almost all traditional image compression standards in terms of rate-distortion (RD) performance. However, the time complexity of LIC model is still underdiscovered, limiting the practical applications in industry. Even with the acceleration of GPU, LIC models still struggle with long coding time, especially on the decoder side. In this paper, we analyze and test a few prevailing and representative LIC models, and compare their complexity with traditional codecs including H.265/HEVC intra and H.266/VVC intra. We provide a comprehensive analysis on every module in the LIC models, and investigate how bitrate changes affect coding time. We observe that the time complexity bottleneck mainly exists in entropy coding and context modelling. Although this paper pay more attention to experimental statistics, our analysis reveals some insights for further acceleration of LIC model, such as model modification for parallel computing, model pruning and a more parallel context model.
Compressed sensing (CS) has recently attracted much interest for its ability to recovery a sparse signal with very limited number of samples. In this paper, we adapt this idea and present a framework of high-resolutio...
详细信息
The following topics are dealt with: video coding; data compression; image coding; convolutional neural nets; decoding; learning (artificial intelligence); motion compensation; video codecs; image reconstruction; filt...
The following topics are dealt with: video coding; data compression; image coding; convolutional neural nets; decoding; learning (artificial intelligence); motion compensation; video codecs; image reconstruction; filtering theory.
a novel single channel SAR-GMTI method is proposed in this paper. When the azimuth mismatch filter performs compression, it induces shifted difference between the stationnry and moving targets because of having ...
详细信息
a novel single channel SAR-GMTI method is proposed in this paper. When the azimuth mismatch filter performs compression, it induces shifted difference between the stationnry and moving targets because of having different Doppler center. The proposed method employs this character to separate moving targets from stationary targets. It can produce two images by pulse compression which make use of two symmetrical mismatch filters in azimuth direction, and then cancel stationary and retain moving targets by subtracting one image from another. Compared with traditional methods, it is applicable to both low and high squint case and the detection capability is significantly improved. The simulation results validate the effectiveness of the proposed method.
Multi-Aspect SAR can obtain rich backscatter information about the illuminated scene. For different kinds of targets in the scene, scattering information related to the geometric characteristics of the target is shown...
详细信息
This paper proposes a novel radar imaging simulator based on analytical electromagnetic and geometric models for a common typical kind of urban structure. The analytical models are presented not only for improving the...
详细信息
In this paper, we propose a novel radar imaging simulator with precise simulation of geometry and relative simulation of radiometry, to assist target recognition. The simulator is based on an new visualization procedu...
详细信息
Interferometric synthetic aperture radar (InSAR) can be used to extract digital elevation model (DEM) with high accuracy. However, the side looking geometry of synthetic aperture radar (SAR) may cause geometric distor...
详细信息
暂无评论