Idris Dulau, M. Beurton-Aimar, Yeykuang Hwu, B. Recur
Field of View Nano-CT X-Ray synchrotron imaging is used for acquiring brain neuronal features from Golgi-stained bio-samples. It theoretically requires a large number of acquired radiographs for compensating reconstruction noise reinforced by the brain features sparsity. However reducing the number of radiographs is essential in routine applications but it results to degraded tomograms. In such a case, traditional segmentation methods are no longer able to distinguish neuronal structures from surrounding noise. We investigate several existing deep-learning networks and we define new ones to segment brain features from very degraded tomograms. We demonstrate the superiority of the proposed networks compared to existing ones.
{"title":"Investigation on Encoder-Decoder Networks for Segmentation of Very Degraded X-Ray CT Tomograms","authors":"Idris Dulau, M. Beurton-Aimar, Yeykuang Hwu, B. Recur","doi":"10.24132/csrn.3301.3","DOIUrl":"https://doi.org/10.24132/csrn.3301.3","url":null,"abstract":"Field of View Nano-CT X-Ray synchrotron imaging is used for acquiring brain neuronal features from Golgi-stained bio-samples. It theoretically requires a large number of acquired radiographs for compensating reconstruction noise reinforced by the brain features sparsity. However reducing the number of radiographs is essential in routine applications but it results to degraded tomograms. In such a case, traditional segmentation methods are no longer able to distinguish neuronal structures from surrounding noise. We investigate several existing deep-learning networks and we define new ones to segment brain features from very degraded tomograms. We demonstrate the superiority of the proposed networks compared to existing ones.","PeriodicalId":322214,"journal":{"name":"Computer Science Research Notes","volume":"58 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121658205","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this paper we deal with the problem of real-time virtual view synthesis, which is crucial in practical immersive video systems. The majority of existing real-time view synthesizers described in literature require using dedicated hardware. In the proposed approach, the view synthesis algorithm is implemented on a CPU increasing its usability for users equipped with consumer devices such as personal computers or laptops. The novelty of the proposed algorithm is based on the atomic z-test function, which allows for parallelization of the depth reprojection step, what was not possible in previous works. The proposal was evaluated on a test set containing miscellaneous perspective and omnidirectional sequences, both in terms of quality and computational time. The results were compared to the state-of-the-art view synthesis algorithm – RVS.
{"title":"Massively Parallel CPU-based Virtual View Synthesis with Atomic Z-test","authors":"J. Stankowski, A. Dziembowski","doi":"10.24132/csrn.3301.32","DOIUrl":"https://doi.org/10.24132/csrn.3301.32","url":null,"abstract":"In this paper we deal with the problem of real-time virtual view synthesis, which is crucial in practical immersive video systems. The majority of existing real-time view synthesizers described in literature require using dedicated hardware. In the proposed approach, the view synthesis algorithm is implemented on a CPU increasing its usability for users equipped with consumer devices such as personal computers or laptops. The novelty of the proposed algorithm is based on the atomic z-test function, which allows for parallelization of the depth reprojection step, what was not possible in previous works. The proposal was evaluated on a test set containing miscellaneous perspective and omnidirectional sequences, both in terms of quality and computational time. The results were compared to the state-of-the-art view synthesis algorithm – RVS.","PeriodicalId":322214,"journal":{"name":"Computer Science Research Notes","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126408303","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ray tracing is typically accelerated by organizing the scene geometry into an acceleration data structure. Hardware-accelerated ray tracing, available through modern graphics APIs, exposes an interface to the acceleration structure (AS) builder that constructs it given the input scene geometry. However, this process is opaque, with limited knowledge and control over the internal algorithm. Additional control is available through the layout of the AS builder input data, the geometry of the scene structured in a user-defined way. In this work, we evaluate the impact of a different scene structuring on the run time performance of the ray-triangle intersections in the context of hardware-accelerated ray tracing. We discuss the possible causes of significantly different outcomes (up to 1.4 times) for the same scene and identify a potential to reduce the cost by automatic input structure optimization.
{"title":"On Importance of Scene Structure for Hardware-Accelerated Ray Tracing","authors":"Martin Káčerik, Jiří Bittner","doi":"10.24132/csrn.3301.60","DOIUrl":"https://doi.org/10.24132/csrn.3301.60","url":null,"abstract":"Ray tracing is typically accelerated by organizing the scene geometry into an acceleration data structure. Hardware-accelerated ray tracing, available through modern graphics APIs, exposes an interface to the acceleration structure (AS) builder that constructs it given the input scene geometry. However, this process is opaque, with limited knowledge and control over the internal algorithm. Additional control is available through the layout of the AS builder input data, the geometry of the scene structured in a user-defined way. In this work, we evaluate the impact of a different scene structuring on the run time performance of the ray-triangle intersections in the context of hardware-accelerated ray tracing. We discuss the possible causes of significantly different outcomes (up to 1.4 times) for the same scene and identify a potential to reduce the cost by automatic input structure optimization.","PeriodicalId":322214,"journal":{"name":"Computer Science Research Notes","volume":"52 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131503131","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The Blocky Volume Package (BVP) format is a distributed, platform-independent and API-independent format for storing static and temporal volumetric data. It is designed for efficient transfer over a network by supporting sparse volumes, multiple resolutions, random access, and streaming, as well as providing a strict framework for supporting a wide palette of encoding formats. The BVP format achieves this by dividing a volume or a volume sequence into blocks that can be compressed and reused. The metadata for the blocks are stored in separate files so that a client has all the information required for loading and decoding the blocks before the actual transmission, decoding and rendering take place. This design allows for random access and parallel loading and has been specifically designed for efficient use on the web platform by adhering to the current living standards. In the paper, we compare the BVP format with some of the most often implemented volume storage formats, and show that the BVP format supports most major features of these formats while at the same time being easily implementable and extensible.
{"title":"Blocky Volume Package: a Web-friendly Volume Storage and Compression Solution","authors":"Žiga Lesar, Ciril Bohak, M. Marolt","doi":"10.24132/csrn.3301.25","DOIUrl":"https://doi.org/10.24132/csrn.3301.25","url":null,"abstract":"The Blocky Volume Package (BVP) format is a distributed, platform-independent and API-independent format for storing static and temporal volumetric data. It is designed for efficient transfer over a network by supporting sparse volumes, multiple resolutions, random access, and streaming, as well as providing a strict framework for supporting a wide palette of encoding formats. The BVP format achieves this by dividing a volume or a volume sequence into blocks that can be compressed and reused. The metadata for the blocks are stored in separate files so that a client has all the information required for loading and decoding the blocks before the actual transmission, decoding and rendering take place. This design allows for random access and parallel loading and has been specifically designed for efficient use on the web platform by adhering to the current living standards. In the paper, we compare the BVP format with some of the most often implemented volume storage formats, and show that the BVP format supports most major features of these formats while at the same time being easily implementable and extensible.","PeriodicalId":322214,"journal":{"name":"Computer Science Research Notes","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126586141","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Automatic analysis of actions in sports training can provide useful feedback for athletes. Fencing is one of the sports disciplines in which the correct technique for performing actions is very important. For any practical application, temporal segmentation of movement in continuous training is crucial. In this work, we consider detecting and classifying actions in a sequence of fencing footwork exercises. We apply pose estimation to RGB videos and then we perform per-frame motion classification, using both classical machine learning and deep learning methods. Using sequences of frames with the same class we find data segments with specific actions. For evaluation, we provide extended manual labels for a fencing footwork dataset previously used in other works. Results indicate that the proposed methods are effective at detecting four footwork actions, obtaining 0.98 F1 score for recognition of action segments and 0.92 F1 score for per-frame classification. In the evaluation of our approach, we provide also a comparison with other data modalities, including depth-based pose estimation and inertial signals. Finally, we include an example of qualitative analysis of the performance of detected actions, to show how this approach can be used for training support.
{"title":"Temporal Segmentation of Actions in Fencing Footwork Training","authors":"F. Malawski, Marek Krupa","doi":"10.24132/csrn.3301.28","DOIUrl":"https://doi.org/10.24132/csrn.3301.28","url":null,"abstract":"Automatic analysis of actions in sports training can provide useful feedback for athletes. Fencing is one of the sports disciplines in which the correct technique for performing actions is very important. For any practical application, temporal segmentation of movement in continuous training is crucial. In this work, we consider detecting and classifying actions in a sequence of fencing footwork exercises. We apply pose estimation to RGB videos and then we perform per-frame motion classification, using both classical machine learning and deep learning methods. Using sequences of frames with the same class we find data segments with specific actions. For evaluation, we provide extended manual labels for a fencing footwork dataset previously used in other works. Results indicate that the proposed methods are effective at detecting four footwork actions, obtaining 0.98 F1 score for recognition of action segments and 0.92 F1 score for per-frame classification. In the evaluation of our approach, we provide also a comparison with other data modalities, including depth-based pose estimation and inertial signals. Finally, we include an example of qualitative analysis of the performance of detected actions, to show how this approach can be used for training support.","PeriodicalId":322214,"journal":{"name":"Computer Science Research Notes","volume":"61 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115933450","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
River generation is an integral part of realistic terrain generation, since rivers shape terrains and changes in terrain, e.g., due to tectonic movements can change the path of rivers. Fast existing terrain generation methods often result in non-realistic river patterns, whereas physically-realistic techniques, e.g., building on erosion models, are usually slow. In this paper we investigate whether the Space Colonization Algorithm can be modified to generate realistic river patterns. We present several extensions of the Space Colonization Algorithm and show with a user study with $n=55$ participants that some variants of the algorithm are capable of generating river patterns that are indistinguishable from real river patterns. Although our technique can not generate all types of natural river patterns, our results suggest that it can prove useful for developing plausible 2D maps and potentially can form the basis for new terrain generation techniques.
{"title":"Generating Realistic River Patterns with Space Colonization","authors":"H. Feng, B. Wünsche, Alex Shaw","doi":"10.24132/csrn.3301.26","DOIUrl":"https://doi.org/10.24132/csrn.3301.26","url":null,"abstract":"River generation is an integral part of realistic terrain generation, since rivers shape terrains and changes in terrain, e.g., due to tectonic movements can change the path of rivers. Fast existing terrain generation methods often result in non-realistic river patterns, whereas physically-realistic techniques, e.g., building on erosion models, are usually slow. In this paper we investigate whether the Space Colonization Algorithm can be modified to generate realistic river patterns. We present several extensions of the Space Colonization Algorithm and show with a user study with $n=55$ participants that some variants of the algorithm are capable of generating river patterns that are indistinguishable from real river patterns. Although our technique can not generate all types of natural river patterns, our results suggest that it can prove useful for developing plausible 2D maps and potentially can form the basis for new terrain generation techniques.","PeriodicalId":322214,"journal":{"name":"Computer Science Research Notes","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121963096","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The problem of recycling secondary raw materials remains unresolved, despite many years of work on this issue. Among the many obstacles that arise is also the difficulty of sorting individual waste fractions. To facilitate this task and help solve this problem, modern computer vision and artificial intelligence techniques can be used. In our work, we propose constructing an intelligent garbage bin containing a camera and a microcomputer along with software that uses these techniques to sort waste. The role of the software is to recognize the type of waste and assign it to one of five main categories: paper, plastic, metal, glass and cardboard. The proposed method uses image recognition techniques with a convolutional neural network. The results confirm that using artificial intelligence methods significantly helps in sorting waste. The proposed solution can be used in homes and public places such as parks, cinemas or playgrounds.
{"title":"The use of Artificial Intelligence for Automatic Waste Segregation in the Garbage Recycling Process","authors":"J. Bobulski, M. Kubanek","doi":"10.24132/csrn.3301.40","DOIUrl":"https://doi.org/10.24132/csrn.3301.40","url":null,"abstract":"The problem of recycling secondary raw materials remains unresolved, despite many years of work on this issue. Among the many obstacles that arise is also the difficulty of sorting individual waste fractions. To facilitate this task and help solve this problem, modern computer vision and artificial intelligence techniques can be used. In our work, we propose constructing an intelligent garbage bin containing a camera and a microcomputer along with software that uses these techniques to sort waste. The role of the software is to recognize the type of waste and assign it to one of five main categories: paper, plastic, metal, glass and cardboard. The proposed method uses image recognition techniques with a convolutional neural network. The results confirm that using artificial intelligence methods significantly helps in sorting waste. The proposed solution can be used in homes and public places such as parks, cinemas or playgrounds.","PeriodicalId":322214,"journal":{"name":"Computer Science Research Notes","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131915335","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
eXpressive B-Spline Curve (XBSC) is a resolution-independent and computationally efficient technique for vector-based stroke modeling and rendering with the flexibility in defining and adjusting the shape and other parameters of the stroke. It generalizes the existing Disk B-Spline Curve (DBSC) geometric representation, which itself is a generalization of the Disk Bézier curve. XBSC allows flexible shape and color manipulation and rendering of strokes with asymmetrical shape control and rich color management. These properties make XBSC suitable for modeling freeform stroke shapes and animation, specifically in squash and stretch, a common technique to bestow elasticity and flexibility in shape changes. During the squash and stretch animation computation, we constrain the shape of the XBSC stroke to conserve its area. To achieve this, we apply the simulated annealing algorithm to iteratively adjust the XBSC while maintaining its area. We show several drawings, rendering and deformation examples to demonstrate the robustness of XBSC.
{"title":"Modeling and Rendering with eXpressive B-Spline Curves","authors":"H. Seah, Budianto Tandianus, Yiliang Sui","doi":"10.24132/csrn.3301.10","DOIUrl":"https://doi.org/10.24132/csrn.3301.10","url":null,"abstract":"eXpressive B-Spline Curve (XBSC) is a resolution-independent and computationally efficient technique for vector-based stroke modeling and rendering with the flexibility in defining and adjusting the shape and other parameters of the stroke. It generalizes the existing Disk B-Spline Curve (DBSC) geometric representation, which itself is a generalization of the Disk Bézier curve. XBSC allows flexible shape and color manipulation and rendering of strokes with asymmetrical shape control and rich color management. These properties make XBSC suitable for modeling freeform stroke shapes and animation, specifically in squash and stretch, a common technique to bestow elasticity and flexibility in shape changes. During the squash and stretch animation computation, we constrain the shape of the XBSC stroke to conserve its area. To achieve this, we apply the simulated annealing algorithm to iteratively adjust the XBSC while maintaining its area. We show several drawings, rendering and deformation examples to demonstrate the robustness of XBSC.","PeriodicalId":322214,"journal":{"name":"Computer Science Research Notes","volume":"104 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122101553","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Angelika Hable, Marko Matore, Anton Scherr, T. Krivec, D. Gruber
The quality inspection of printed circuit boards (PCBs) is no longer feasible by human inspectors due to accuracy requirements and the processing volume. Automated optical inspection systems must be specifically designed to meet the various inspection requirements. A photometric stereo setup is suitable for the inspection of highly reflective gold areas on PCBs. In this process, several image acquisitions are performed under different illumination directions. This can reveal defects that are not visible under other illumination systems. In this paper, we use a segmented ring light so that image acquisition is possible under four different illumination directions. Using these images, a normal map and a mean image are calculated. The defects on the gold areas of PCBs are detectable in either the normal map, the mean image, or both images. A CNN for classification detects the defects. The input is a 6-dimensional image from normal map and mean image. An accuracy up to 95% can be achieved with the available dataset.
{"title":"Detection of Printed Circuit Board Defects with Photometric Stereo and Convolutional Neural Networks","authors":"Angelika Hable, Marko Matore, Anton Scherr, T. Krivec, D. Gruber","doi":"10.24132/csrn.3301.92","DOIUrl":"https://doi.org/10.24132/csrn.3301.92","url":null,"abstract":"The quality inspection of printed circuit boards (PCBs) is no longer feasible by human inspectors due to accuracy requirements and the processing volume. Automated optical inspection systems must be specifically designed to meet the various inspection requirements. A photometric stereo setup is suitable for the inspection of highly reflective gold areas on PCBs. In this process, several image acquisitions are performed under different illumination directions. This can reveal defects that are not visible under other illumination systems. In this paper, we use a segmented ring light so that image acquisition is possible under four different illumination directions. Using these images, a normal map and a mean image are calculated. The defects on the gold areas of PCBs are detectable in either the normal map, the mean image, or both images. A CNN for classification detects the defects. The input is a 6-dimensional image from normal map and mean image. An accuracy up to 95% can be achieved with the available dataset.","PeriodicalId":322214,"journal":{"name":"Computer Science Research Notes","volume":"54 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128679581","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We propose a rendering method for efficiently computing the transmitted caustics produced by a glass panel with arbitrary surface deformations, characteristic of old glass used in 3D reconstructions in virtual heritage. Using Fermat"s principle of least time, we generalize the concept of Next Event Estimation to allow light sampling through two displaced refractive interfaces, which amount to numerically finding all stationary points of an objective function. Our work allows for an efficient estimation of the caustic while staying inside a standard Monte Carlo pathtracing framework. Our specific geometrical context allows our solver to converge significantly faster than the more general method Specular Manifold Sampling, while scaling well with the number of panels present in the scene.
{"title":"Optimised Light Rendering through Old Glass","authors":"Quentin Huan, F. Rousselle, C. Renaud","doi":"10.24132/csrn.3301.30","DOIUrl":"https://doi.org/10.24132/csrn.3301.30","url":null,"abstract":"We propose a rendering method for efficiently computing the transmitted caustics produced by a glass panel with arbitrary surface deformations, characteristic of old glass used in 3D reconstructions in virtual heritage. Using Fermat\"s principle of least time, we generalize the concept of Next Event Estimation to allow light sampling through two displaced refractive interfaces, which amount to numerically finding all stationary points of an objective function. Our work allows for an efficient estimation of the caustic while staying inside a standard Monte Carlo pathtracing framework. Our specific geometrical context allows our solver to converge significantly faster than the more general method Specular Manifold Sampling, while scaling well with the number of panels present in the scene.","PeriodicalId":322214,"journal":{"name":"Computer Science Research Notes","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126481433","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：