Pub Date : 2022-11-29DOI: 10.48550/arXiv.2211.16066
Bram Vanherle, Steven Moonen, F. Reeth, Nick Michiels
Recently, the use of synthetic training data has been on the rise as it offers correctly labelled datasets at a lower cost. The downside of this technique is that the so-called domain gap between the real target images and synthetic training data leads to a decrease in performance. In this paper, we attempt to provide a holistic overview of how to use synthetic data for object detection. We analyse aspects of generating the data as well as techniques used to train the models. We do so by devising a number of experiments, training models on the Dataset of Industrial Metal Objects (DIMO). This dataset contains both real and synthetic images. The synthetic part has different subsets that are either exact synthetic copies of the real data or are copies with certain aspects randomised. This allows us to analyse what types of variation are good for synthetic training data and which aspects should be modelled to closely match the target data. Furthermore, we investigate what types of training techniques are beneficial towards generalisation to real data, and how to use them. Additionally, we analyse how real images can be leveraged when training on synthetic images. All these experiments are validated on real data and benchmarked to models trained on real data. The results offer a number of interesting takeaways that can serve as basic guidelines for using synthetic data for object detection. Code to reproduce results is available at https://github.com/EDM-Research/DIMO_ObjectDetection.
{"title":"Analysis of Training Object Detection Models with Synthetic Data","authors":"Bram Vanherle, Steven Moonen, F. Reeth, Nick Michiels","doi":"10.48550/arXiv.2211.16066","DOIUrl":"https://doi.org/10.48550/arXiv.2211.16066","url":null,"abstract":"Recently, the use of synthetic training data has been on the rise as it offers correctly labelled datasets at a lower cost. The downside of this technique is that the so-called domain gap between the real target images and synthetic training data leads to a decrease in performance. In this paper, we attempt to provide a holistic overview of how to use synthetic data for object detection. We analyse aspects of generating the data as well as techniques used to train the models. We do so by devising a number of experiments, training models on the Dataset of Industrial Metal Objects (DIMO). This dataset contains both real and synthetic images. The synthetic part has different subsets that are either exact synthetic copies of the real data or are copies with certain aspects randomised. This allows us to analyse what types of variation are good for synthetic training data and which aspects should be modelled to closely match the target data. Furthermore, we investigate what types of training techniques are beneficial towards generalisation to real data, and how to use them. Additionally, we analyse how real images can be leveraged when training on synthetic images. All these experiments are validated on real data and benchmarked to models trained on real data. The results offer a number of interesting takeaways that can serve as basic guidelines for using synthetic data for object detection. Code to reproduce results is available at https://github.com/EDM-Research/DIMO_ObjectDetection.","PeriodicalId":72437,"journal":{"name":"BMVC : proceedings of the British Machine Vision Conference. British Machine Vision Conference","volume":"252 1","pages":"833"},"PeriodicalIF":0.0,"publicationDate":"2022-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78199662","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}
Pub Date : 2022-11-25DOI: 10.48550/arXiv.2211.14005
M. Nottebaum, S. Roth, Simone Schaub-Meyer
Most deep learning methods for video frame interpolation consist of three main components: feature extraction, motion estimation, and image synthesis. Existing approaches are mainly distinguishable in terms of how these modules are designed. However, when interpolating high-resolution images, e.g. at 4K, the design choices for achieving high accuracy within reasonable memory requirements are limited. The feature extraction layers help to compress the input and extract relevant information for the latter stages, such as motion estimation. However, these layers are often costly in parameters, computation time, and memory. We show how ideas from dimensionality reduction combined with a lightweight optimization can be used to compress the input representation while keeping the extracted information suitable for frame interpolation. Further, we require neither a pretrained flow network nor a synthesis network, additionally reducing the number of trainable parameters and required memory. When evaluating on three 4K benchmarks, we achieve state-of-the-art image quality among the methods without pretrained flow while having the lowest network complexity and memory requirements overall.
{"title":"Efficient Feature Extraction for High-resolution Video Frame Interpolation","authors":"M. Nottebaum, S. Roth, Simone Schaub-Meyer","doi":"10.48550/arXiv.2211.14005","DOIUrl":"https://doi.org/10.48550/arXiv.2211.14005","url":null,"abstract":"Most deep learning methods for video frame interpolation consist of three main components: feature extraction, motion estimation, and image synthesis. Existing approaches are mainly distinguishable in terms of how these modules are designed. However, when interpolating high-resolution images, e.g. at 4K, the design choices for achieving high accuracy within reasonable memory requirements are limited. The feature extraction layers help to compress the input and extract relevant information for the latter stages, such as motion estimation. However, these layers are often costly in parameters, computation time, and memory. We show how ideas from dimensionality reduction combined with a lightweight optimization can be used to compress the input representation while keeping the extracted information suitable for frame interpolation. Further, we require neither a pretrained flow network nor a synthesis network, additionally reducing the number of trainable parameters and required memory. When evaluating on three 4K benchmarks, we achieve state-of-the-art image quality among the methods without pretrained flow while having the lowest network complexity and memory requirements overall.","PeriodicalId":72437,"journal":{"name":"BMVC : proceedings of the British Machine Vision Conference. British Machine Vision Conference","volume":"24 1","pages":"825"},"PeriodicalIF":0.0,"publicationDate":"2022-11-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91092339","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}
Pub Date : 2022-11-25DOI: 10.48550/arXiv.2211.14037
R. Groenendijk, L. Dorst, T. Gevers
Pooling is essentially an operation from the field of Mathematical Morphology, with max pooling as a limited special case. The more general setting of MorphPooling greatly extends the tool set for building neural networks. In addition to pooling operations, encoder-decoder networks used for pixel-level predictions also require unpooling. It is common to combine unpooling with convolution or deconvolution for up-sampling. However, using its morphological properties, unpooling can be generalised and improved. Extensive experimentation on two tasks and three large-scale datasets shows that morphological pooling and unpooling lead to improved predictive performance at much reduced parameter counts.
{"title":"MorphPool: Efficient Non-linear Pooling & Unpooling in CNNs","authors":"R. Groenendijk, L. Dorst, T. Gevers","doi":"10.48550/arXiv.2211.14037","DOIUrl":"https://doi.org/10.48550/arXiv.2211.14037","url":null,"abstract":"Pooling is essentially an operation from the field of Mathematical Morphology, with max pooling as a limited special case. The more general setting of MorphPooling greatly extends the tool set for building neural networks. In addition to pooling operations, encoder-decoder networks used for pixel-level predictions also require unpooling. It is common to combine unpooling with convolution or deconvolution for up-sampling. However, using its morphological properties, unpooling can be generalised and improved. Extensive experimentation on two tasks and three large-scale datasets shows that morphological pooling and unpooling lead to improved predictive performance at much reduced parameter counts.","PeriodicalId":72437,"journal":{"name":"BMVC : proceedings of the British Machine Vision Conference. British Machine Vision Conference","volume":"21 1","pages":"56"},"PeriodicalIF":0.0,"publicationDate":"2022-11-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78470486","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}
Pub Date : 2022-11-25DOI: 10.48550/arXiv.2211.14074
Liang Zeng, A. Lengyel, Nergis Tomen, J. V. Gemert
In this work, we leverage estimated depth to boost self-supervised contrastive learning for segmentation of urban scenes, where unlabeled videos are readily available for training self-supervised depth estimation. We argue that the semantics of a coherent group of pixels in 3D space is self-contained and invariant to the contexts in which they appear. We group coherent, semantically related pixels into coherent depth regions given their estimated depth and use copy-paste to synthetically vary their contexts. In this way, cross-context correspondences are built in contrastive learning and a context-invariant representation is learned. For unsupervised semantic segmentation of urban scenes, our method surpasses the previous state-of-the-art baseline by +7.14% in mIoU on Cityscapes and +6.65% on KITTI. For fine-tuning on Cityscapes and KITTI segmentation, our method is competitive with existing models, yet, we do not need to pre-train on ImageNet or COCO, and we are also more computationally efficient. Our code is available on https://github.com/LeungTsang/CPCDR
{"title":"Copy-Pasting Coherent Depth Regions Improves Contrastive Learning for Urban-Scene Segmentation","authors":"Liang Zeng, A. Lengyel, Nergis Tomen, J. V. Gemert","doi":"10.48550/arXiv.2211.14074","DOIUrl":"https://doi.org/10.48550/arXiv.2211.14074","url":null,"abstract":"In this work, we leverage estimated depth to boost self-supervised contrastive learning for segmentation of urban scenes, where unlabeled videos are readily available for training self-supervised depth estimation. We argue that the semantics of a coherent group of pixels in 3D space is self-contained and invariant to the contexts in which they appear. We group coherent, semantically related pixels into coherent depth regions given their estimated depth and use copy-paste to synthetically vary their contexts. In this way, cross-context correspondences are built in contrastive learning and a context-invariant representation is learned. For unsupervised semantic segmentation of urban scenes, our method surpasses the previous state-of-the-art baseline by +7.14% in mIoU on Cityscapes and +6.65% on KITTI. For fine-tuning on Cityscapes and KITTI segmentation, our method is competitive with existing models, yet, we do not need to pre-train on ImageNet or COCO, and we are also more computationally efficient. Our code is available on https://github.com/LeungTsang/CPCDR","PeriodicalId":72437,"journal":{"name":"BMVC : proceedings of the British Machine Vision Conference. British Machine Vision Conference","volume":"75 1","pages":"893"},"PeriodicalIF":0.0,"publicationDate":"2022-11-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83779256","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}
Pub Date : 2022-11-24DOI: 10.48550/arXiv.2211.13429
Ziwei Yu, Linlin Yang, You Xie, Ping Chen, Angela Yao
We propose a novel framework for 3D hand shape reconstruction and hand-object grasp optimization from a single RGB image. The representation of hand-object contact regions is critical for accurate reconstructions. Instead of approximating the contact regions with sparse points, as in previous works, we propose a dense representation in the form of a UV coordinate map. Furthermore, we introduce inference-time optimization to fine-tune the grasp and improve interactions between the hand and the object. Our pipeline increases hand shape reconstruction accuracy and produces a vibrant hand texture. Experiments on datasets such as Ho3D, FreiHAND, and DexYCB reveal that our proposed method outperforms the state-of-the-art.
{"title":"UV-Based 3D Hand-Object Reconstruction with Grasp Optimization","authors":"Ziwei Yu, Linlin Yang, You Xie, Ping Chen, Angela Yao","doi":"10.48550/arXiv.2211.13429","DOIUrl":"https://doi.org/10.48550/arXiv.2211.13429","url":null,"abstract":"We propose a novel framework for 3D hand shape reconstruction and hand-object grasp optimization from a single RGB image. The representation of hand-object contact regions is critical for accurate reconstructions. Instead of approximating the contact regions with sparse points, as in previous works, we propose a dense representation in the form of a UV coordinate map. Furthermore, we introduce inference-time optimization to fine-tune the grasp and improve interactions between the hand and the object. Our pipeline increases hand shape reconstruction accuracy and produces a vibrant hand texture. Experiments on datasets such as Ho3D, FreiHAND, and DexYCB reveal that our proposed method outperforms the state-of-the-art.","PeriodicalId":72437,"journal":{"name":"BMVC : proceedings of the British Machine Vision Conference. British Machine Vision Conference","volume":"46 1","pages":"111"},"PeriodicalIF":0.0,"publicationDate":"2022-11-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74051699","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}
Pub Date : 2022-11-24DOI: 10.48550/arXiv.2211.13465
Otabek Nazarov, Mohammad Yaqub, K. Nandakumar
Chest X-ray is one of the most popular medical imaging modalities due to its accessibility and effectiveness. However, there is a chronic shortage of well-trained radiologists who can interpret these images and diagnose the patient's condition. Therefore, automated radiology report generation can be a very helpful tool in clinical practice. A typical report generation workflow consists of two main steps: (i) encoding the image into a latent space and (ii) generating the text of the report based on the latent image embedding. Many existing report generation techniques use a standard convolutional neural network (CNN) architecture for image encoding followed by a Transformer-based decoder for medical text generation. In most cases, CNN and the decoder are trained jointly in an end-to-end fashion. In this work, we primarily focus on understanding the relative importance of encoder and decoder components. Towards this end, we analyze four different image encoding approaches: direct, fine-grained, CLIP-based, and Cluster-CLIP-based encodings in conjunction with three different decoders on the large-scale MIMIC-CXR dataset. Among these encoders, the cluster CLIP visual encoder is a novel approach that aims to generate more discriminative and explainable representations. CLIP-based encoders produce comparable results to traditional CNN-based encoders in terms of NLP metrics, while fine-grained encoding outperforms all other encoders both in terms of NLP and clinical accuracy metrics, thereby validating the importance of image encoder to effectively extract semantic information. GitHub repository: https://github.com/mudabek/encoding-cxr-report-gen
{"title":"On the Importance of Image Encoding in Automated Chest X-Ray Report Generation","authors":"Otabek Nazarov, Mohammad Yaqub, K. Nandakumar","doi":"10.48550/arXiv.2211.13465","DOIUrl":"https://doi.org/10.48550/arXiv.2211.13465","url":null,"abstract":"Chest X-ray is one of the most popular medical imaging modalities due to its accessibility and effectiveness. However, there is a chronic shortage of well-trained radiologists who can interpret these images and diagnose the patient's condition. Therefore, automated radiology report generation can be a very helpful tool in clinical practice. A typical report generation workflow consists of two main steps: (i) encoding the image into a latent space and (ii) generating the text of the report based on the latent image embedding. Many existing report generation techniques use a standard convolutional neural network (CNN) architecture for image encoding followed by a Transformer-based decoder for medical text generation. In most cases, CNN and the decoder are trained jointly in an end-to-end fashion. In this work, we primarily focus on understanding the relative importance of encoder and decoder components. Towards this end, we analyze four different image encoding approaches: direct, fine-grained, CLIP-based, and Cluster-CLIP-based encodings in conjunction with three different decoders on the large-scale MIMIC-CXR dataset. Among these encoders, the cluster CLIP visual encoder is a novel approach that aims to generate more discriminative and explainable representations. CLIP-based encoders produce comparable results to traditional CNN-based encoders in terms of NLP metrics, while fine-grained encoding outperforms all other encoders both in terms of NLP and clinical accuracy metrics, thereby validating the importance of image encoder to effectively extract semantic information. GitHub repository: https://github.com/mudabek/encoding-cxr-report-gen","PeriodicalId":72437,"journal":{"name":"BMVC : proceedings of the British Machine Vision Conference. British Machine Vision Conference","volume":"45 1","pages":"475"},"PeriodicalIF":0.0,"publicationDate":"2022-11-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88629050","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}
Three-dimensional (3D) object reconstruction based on differentiable rendering (DR) is an active research topic in computer vision. DR-based methods minimize the difference between the rendered and target images by optimizing both the shape and appearance and realizing a high visual reproductivity. However, most approaches perform poorly for textureless objects because of the geometrical ambiguity, which means that multiple shapes can have the same rendered result in such objects. To overcome this problem, we introduce active sensing with structured light (SL) into multi-view 3D object reconstruction based on DR to learn the unknown geometry and appearance of arbitrary scenes and camera poses. More specifically, our framework leverages the correspondences between pixels in different views calculated by structured light as an additional constraint in the DR-based optimization of implicit surface, color representations, and camera poses. Because camera poses can be optimized simultaneously, our method realizes high reconstruction accuracy in the textureless region and reduces efforts for camera pose calibration, which is required for conventional SL-based methods. Experiment results on both synthetic and real data demonstrate that our system outperforms conventional DR- and SL-based methods in a high-quality surface reconstruction, particularly for challenging objects with textureless or shiny surfaces.
{"title":"Multi-View Neural Surface Reconstruction with Structured Light","authors":"Chunyu Li, Taisuke Hashimoto, Eiichi Matsumoto, Hiroharu Kato","doi":"10.48550/arXiv.2211.11971","DOIUrl":"https://doi.org/10.48550/arXiv.2211.11971","url":null,"abstract":"Three-dimensional (3D) object reconstruction based on differentiable rendering (DR) is an active research topic in computer vision. DR-based methods minimize the difference between the rendered and target images by optimizing both the shape and appearance and realizing a high visual reproductivity. However, most approaches perform poorly for textureless objects because of the geometrical ambiguity, which means that multiple shapes can have the same rendered result in such objects. To overcome this problem, we introduce active sensing with structured light (SL) into multi-view 3D object reconstruction based on DR to learn the unknown geometry and appearance of arbitrary scenes and camera poses. More specifically, our framework leverages the correspondences between pixels in different views calculated by structured light as an additional constraint in the DR-based optimization of implicit surface, color representations, and camera poses. Because camera poses can be optimized simultaneously, our method realizes high reconstruction accuracy in the textureless region and reduces efforts for camera pose calibration, which is required for conventional SL-based methods. Experiment results on both synthetic and real data demonstrate that our system outperforms conventional DR- and SL-based methods in a high-quality surface reconstruction, particularly for challenging objects with textureless or shiny surfaces.","PeriodicalId":72437,"journal":{"name":"BMVC : proceedings of the British Machine Vision Conference. British Machine Vision Conference","volume":"31 1","pages":"550"},"PeriodicalIF":0.0,"publicationDate":"2022-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91483064","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}
Pub Date : 2022-11-22DOI: 10.48550/arXiv.2211.12209
Krishnakant Singh, Simone Schaub-Meyer, S. Roth
The high-quality images yielded by generative adversarial networks (GANs) have motivated investigations into their application for image editing. However, GANs are often limited in the control they provide for performing specific edits. One of the principal challenges is the entangled latent space of GANs, which is not directly suitable for performing independent and detailed edits. Recent editing methods allow for either controlled style edits or controlled semantic edits. In addition, methods that use semantic masks to edit images have difficulty preserving the identity and are unable to perform controlled style edits. We propose a method to disentangle a GAN$text{'}$s latent space into semantic and style spaces, enabling controlled semantic and style edits for face images independently within the same framework. To achieve this, we design an encoder-decoder based network architecture ($S^2$-Flow), which incorporates two proposed inductive biases. We show the suitability of $S^2$-Flow quantitatively and qualitatively by performing various semantic and style edits.
{"title":"S2-Flow: Joint Semantic and Style Editing of Facial Images","authors":"Krishnakant Singh, Simone Schaub-Meyer, S. Roth","doi":"10.48550/arXiv.2211.12209","DOIUrl":"https://doi.org/10.48550/arXiv.2211.12209","url":null,"abstract":"The high-quality images yielded by generative adversarial networks (GANs) have motivated investigations into their application for image editing. However, GANs are often limited in the control they provide for performing specific edits. One of the principal challenges is the entangled latent space of GANs, which is not directly suitable for performing independent and detailed edits. Recent editing methods allow for either controlled style edits or controlled semantic edits. In addition, methods that use semantic masks to edit images have difficulty preserving the identity and are unable to perform controlled style edits. We propose a method to disentangle a GAN$text{'}$s latent space into semantic and style spaces, enabling controlled semantic and style edits for face images independently within the same framework. To achieve this, we design an encoder-decoder based network architecture ($S^2$-Flow), which incorporates two proposed inductive biases. We show the suitability of $S^2$-Flow quantitatively and qualitatively by performing various semantic and style edits.","PeriodicalId":72437,"journal":{"name":"BMVC : proceedings of the British Machine Vision Conference. British Machine Vision Conference","volume":"1 1","pages":"821"},"PeriodicalIF":0.0,"publicationDate":"2022-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88216275","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}
Pub Date : 2022-11-21DOI: 10.48550/arXiv.2211.11131
Jong-Lyul Jeong, Jong-Hwan Kim
Road scene understanding tasks have recently become crucial for self-driving vehicles. In particular, real-time semantic segmentation is indispensable for intelligent self-driving agents to recognize roadside objects in the driving area. As prior research works have primarily sought to improve the segmentation performance with computationally heavy operations, they require far significant hardware resources for both training and deployment, and thus are not suitable for real-time applications. As such, we propose a doubly contrastive approach to improve the performance of a more practical lightweight model for self-driving, specifically under adverse weather conditions such as fog, nighttime, rain and snow. Our proposed approach exploits both image- and pixel-level contrasts in an end-to-end supervised learning scheme without requiring a memory bank for global consistency or the pretraining step used in conventional contrastive methods. We validate the effectiveness of our method using SwiftNet on the ACDC dataset, where it achieves up to 1.34%p improvement in mIoU (ResNet-18 backbone) at 66.7 FPS (2048x1024 resolution) on a single RTX 3080 Mobile GPU at inference. Furthermore, we demonstrate that replacing image-level supervision with self-supervision achieves comparable performance when pre-trained with clear weather images.
{"title":"Doubly Contrastive End-to-End Semantic Segmentation for Autonomous Driving under Adverse Weather","authors":"Jong-Lyul Jeong, Jong-Hwan Kim","doi":"10.48550/arXiv.2211.11131","DOIUrl":"https://doi.org/10.48550/arXiv.2211.11131","url":null,"abstract":"Road scene understanding tasks have recently become crucial for self-driving vehicles. In particular, real-time semantic segmentation is indispensable for intelligent self-driving agents to recognize roadside objects in the driving area. As prior research works have primarily sought to improve the segmentation performance with computationally heavy operations, they require far significant hardware resources for both training and deployment, and thus are not suitable for real-time applications. As such, we propose a doubly contrastive approach to improve the performance of a more practical lightweight model for self-driving, specifically under adverse weather conditions such as fog, nighttime, rain and snow. Our proposed approach exploits both image- and pixel-level contrasts in an end-to-end supervised learning scheme without requiring a memory bank for global consistency or the pretraining step used in conventional contrastive methods. We validate the effectiveness of our method using SwiftNet on the ACDC dataset, where it achieves up to 1.34%p improvement in mIoU (ResNet-18 backbone) at 66.7 FPS (2048x1024 resolution) on a single RTX 3080 Mobile GPU at inference. Furthermore, we demonstrate that replacing image-level supervision with self-supervision achieves comparable performance when pre-trained with clear weather images.","PeriodicalId":72437,"journal":{"name":"BMVC : proceedings of the British Machine Vision Conference. British Machine Vision Conference","volume":"2 1","pages":"460"},"PeriodicalIF":0.0,"publicationDate":"2022-11-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79144144","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}
Pub Date : 2022-11-21DOI: 10.48550/arXiv.2211.11386
Satoshi Ikehata
Existing deep calibrated photometric stereo networks basically aggregate observations under different lights based on the pre-defined operations such as linear projection and max pooling. While they are effective with the dense capture, simple first-order operations often fail to capture the high-order interactions among observations under small number of different lights. To tackle this issue, this paper presents a deep sparse calibrated photometric stereo network named {it PS-Transformer} which leverages the learnable self-attention mechanism to properly capture the complex inter-image interactions. PS-Transformer builds upon the dual-branch design to explore both pixel-wise and image-wise features and individual feature is trained with the intermediate surface normal supervision to maximize geometric feasibility. A new synthetic dataset named CyclesPS+ is also presented with the comprehensive analysis to successfully train the photometric stereo networks. Extensive results on the publicly available benchmark datasets demonstrate that the surface normal prediction accuracy of the proposed method significantly outperforms other state-of-the-art algorithms with the same number of input images and is even comparable to that of dense algorithms which input 10$times$ larger number of images.
{"title":"PS-Transformer: Learning Sparse Photometric Stereo Network using Self-Attention Mechanism","authors":"Satoshi Ikehata","doi":"10.48550/arXiv.2211.11386","DOIUrl":"https://doi.org/10.48550/arXiv.2211.11386","url":null,"abstract":"Existing deep calibrated photometric stereo networks basically aggregate observations under different lights based on the pre-defined operations such as linear projection and max pooling. While they are effective with the dense capture, simple first-order operations often fail to capture the high-order interactions among observations under small number of different lights. To tackle this issue, this paper presents a deep sparse calibrated photometric stereo network named {it PS-Transformer} which leverages the learnable self-attention mechanism to properly capture the complex inter-image interactions. PS-Transformer builds upon the dual-branch design to explore both pixel-wise and image-wise features and individual feature is trained with the intermediate surface normal supervision to maximize geometric feasibility. A new synthetic dataset named CyclesPS+ is also presented with the comprehensive analysis to successfully train the photometric stereo networks. Extensive results on the publicly available benchmark datasets demonstrate that the surface normal prediction accuracy of the proposed method significantly outperforms other state-of-the-art algorithms with the same number of input images and is even comparable to that of dense algorithms which input 10$times$ larger number of images.","PeriodicalId":72437,"journal":{"name":"BMVC : proceedings of the British Machine Vision Conference. British Machine Vision Conference","volume":"11 1","pages":"30"},"PeriodicalIF":0.0,"publicationDate":"2022-11-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87630751","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：