Filters








41,297 Hits in 4.7 sec

Instance Segmentation and Tracking with Cosine Embeddings and Recurrent Hourglass Networks [article]

Christian Payer, Darko Štern, Thomas Neff, Horst Bischof, Martin Urschler
2018 arXiv   pre-print
Afterwards, these embeddings are clustered among subsequent video frames to create the final tracked instance segmentations.  ...  Furthermore, we show applicability of the cosine embedding loss for segmenting leaf instances on still images of plants.  ...  As the number of instances is not known, we perform this grouping with the clustering algorithm HDBSCAN [3] that estimates the number of clusters automatically.  ... 
arXiv:1806.02070v3 fatcat:xyjzrqo6sfcr3bowmbknibkdae

Learning a Spatio-Temporal Embedding for Video Instance Segmentation [article]

Anthony Hu, Alex Kendall, Roberto Cipolla
2019 arXiv   pre-print
In this embedding space, video-pixels of the same instance are clustered together while being separated from other instances, to naturally track instances over time without any complex post-processing.  ...  We show that our model can accurately track and segment instances, even with occlusions and missed detections, advancing the state-of-the-art on the KITTI Multi-Object and Tracking Dataset.  ...  The embedding is visualised in 2D and coloured with the results of the mean shift clustering.  ... 
arXiv:1912.08969v1 fatcat:wvyylpvju5go5iqe7qwxyu6chu

Single-Image Piece-Wise Planar 3D Reconstruction via Associative Embedding

Zehao Yu, Jia Zheng, Dongze Lian, Zihan Zhou, Shenghua Gao
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
However, these methods are limited to detecting a fixed number of planes with certain learned order.  ...  Then, the plane instances are obtained by grouping the embedding vectors in planar regions via an efficient mean shift clustering algorithm.  ...  of clusters C (planes) in the ground truth, N c is the number of elements in cluster c, x i is the pixel embedding, µ c is the mean embedding of the cluster c, and δ v and δ d are the margin for "pull  ... 
doi:10.1109/cvpr.2019.00112 dblp:conf/cvpr/YuZLZG19 fatcat:ei34v24qkrgnbeep6drhbupgje

OccuSeg: Occupancy-Aware 3D Instance Segmentation

Lei Han, Tian Zheng, Lan Xu, Lu Fang
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
The occupancy signal is learnt jointly with a combination of feature and spatial embedding and employed to guide the clustering stage of 3D instance segmentation.  ...  [30] introduce a learnable clustering bandwidth instead of learning embedding using hand-crafted cost functions, achieving accurate instance segmentation in real-time.  ... 
doi:10.1109/cvpr42600.2020.00301 dblp:conf/cvpr/HanZXF20 fatcat:bolm3hjflzeptkd5yv45l7eafe

OccuSeg: Occupancy-aware 3D Instance Segmentation [article]

Lei Han, Tian Zheng, Lan Xu, Lu Fang
2020 arXiv   pre-print
3D instance segmentation, with a variety of applications in robotics and augmented reality, is in large demands these days.  ...  Our multi-task learning produces both occupancy signal and embedding representations, where the training of spatial and feature embeddings varies with their difference in scale-aware.  ...  The occupancy signal is learnt jointly with a combination of feature and spatial embedding and employed to guide the clustering stage of 3D instance segmentation.  ... 
arXiv:2003.06537v3 fatcat:gj3m2eqjevhrldw2t7u7vxjpv4

Single-Image Piece-wise Planar 3D Reconstruction via Associative Embedding [article]

Zehao Yu and Jia Zheng and Dongze Lian and Zihan Zhou and Shenghua Gao
2019 arXiv   pre-print
However, these methods are limited to detecting a fixed number of planes with certain learned order.  ...  Then, the plane instances are obtained by grouping the embedding vectors in planar regions via an efficient mean shift clustering algorithm.  ...  of clusters C (planes) in the ground truth, N c is the number of elements in cluster c, x i is the pixel embedding, µ c is the mean embedding of the cluster c, and δ v and δ d are the margin for "pull  ... 
arXiv:1902.09777v3 fatcat:fpego7f2lvfmjnhr4jrigukhm4

Improving Pixel Embedding Learning through Intermediate Distance Regression Supervision for Instance Segmentation [article]

Yuli Wu, Long Chen, Dorit Merhof
2020 arXiv   pre-print
As a proposal-free approach, instance segmentation through pixel embedding learning and clustering is gaining more emphasis.  ...  By simply concatenating features of the distance regression module to the images as inputs of the embedding module, the mSBD scores on the CVPPP Leaf Segmentation Challenge can be further improved by more  ...  Learning object-aware pixel embeddings is one of the trends in the field of instance segmentation. The embedding is essentially a high-dimensional representation of each pixel.  ... 
arXiv:2007.06660v1 fatcat:ccmqyu55gjfbthc2s7mqhgwjli

CPSeg: Cluster-free Panoptic Segmentation of 3D LiDAR Point Clouds [article]

Enxu Li, Ryan Razani, Yixuan Xu, Bingbing Liu
2021 arXiv   pre-print
Moreover, CPSeg incorporates a new cluster-free instance segmentation head to dynamically pillarize foreground points according to the learned embedding.  ...  Thus, the conventional proposal-based or clustering-based instance segmentation is transformed into a binary segmentation problem on the pairwise embedding comparison matrix.  ...  These improvements can be mainly attributed to the use of cluster-free instance segmentation module and the incorporation of surface normal as a helpful part of the instance embedding.  ... 
arXiv:2111.01723v1 fatcat:e2ymumzyx5grdnqgngqap7tlxq

Probabilistic Deep Learning for Instance Segmentation [article]

Josef Lorenz Rumberger, Lisa Mais, Dagmar Kainmueller
2020 arXiv   pre-print
Furthermore, we analyze the quality of the uncertainty estimates with a metric adapted from semantic segmentation.  ...  These were used in active learning frameworks to target the labeling efforts of specialist annotators or to assess the quality of a prediction in a safety-critical environment.  ...  Post-Processing To yield an instance segmentation from predictions of embeddings, mean-shift clustering [5] is applied to find cluster centers in the embedding space.  ... 
arXiv:2008.10678v2 fatcat:whgu7jjmxfhn5ezbfcy3ebchjy

Sparse Object-level Supervision for Instance Segmentation with Pixel Embeddings [article]

Adrian Wolny, Qin Yu, Constantin Pape, Anna Kreshuk
2022 arXiv   pre-print
We propose to address the dense annotation bottleneck by introducing a proposal-free segmentation approach based on non-spatial embeddings, which exploits the structure of the learned embedding space to  ...  We evaluate the proposed method on 2D and 3D segmentation problems in different microscopy modalities as well as on the Cityscapes and CVPPP instance segmentation benchmarks, achieving state-of-the-art  ...  number of labeled clusters/instances and N U is the number of pixels in the unlabeled region U .  ... 
arXiv:2103.14572v3 fatcat:xd53ngvpxvawpkr3prxomq4cmy

On the effectiveness of feature set augmentation using clusters of word embeddings [article]

Georgios Balikas, Ioannis Partalas
2018 arXiv   pre-print
Despite their importance, their incorporation in the standard pipeline of feature engineering relies more on a trial-and-error procedure where one evaluates several hyper-parameters, like the number of  ...  In order to better understand the role of such features we systematically evaluate their effect on four tasks, those of named entity segmentation and classification as well as, those of five-point sentiment  ...  word-vectors. 2 We cluster the embeddings with k-Means.  ... 
arXiv:1705.01265v2 fatcat:m6fwfunbqfeozc3gqnngvh5sq4

Semantic Instance Segmentation with a Discriminative Loss Function [article]

Bert De Brabandere, Davy Neven, Luc Van Gool
2017 arXiv   pre-print
Our approach of combining an off-the-shelf network with a principled loss function inspired by a metric learning objective is conceptually simple and distinct from recent efforts in instance segmentation  ...  easily be clustered into instances with a simple post-processing step.  ...  Acknowledgement: This work was supported by Toyota, and was carried out at the TRACE Lab at KU Leuven (Toyota Research on Automated Cars in Europe -Leuven).  ... 
arXiv:1708.02551v1 fatcat:lfxqer5fffbw3dmdpc4nxcyohm

Multi-view Incremental Segmentation of 3D Point Clouds for Mobile Robots

Jingdao Chen, Yong Kwon Cho, Zsolt Kira
2019 IEEE Robotics and Automation Letters  
Existing techniques for point cloud semantic segmentation are mostly applied on a single-frame or offline basis, with no way to integrate the segmentation results over time.  ...  The proposed deep neural network, MCPNet, is trained to predict class labels and object instance labels for each point in the scanned point cloud in an incremental fashion.  ...  The computed feature embedding is used to cluster the new points together with previous object instances.  ... 
doi:10.1109/lra.2019.2894915 fatcat:zgbpoxfx3vd2jmsrrggbkroyvi

Learning Shape Segmentation Using Constrained Spectral Clustering and Probabilistic Label Transfer [chapter]

Avinash Sharma, Etienne von Lavante, Radu Horaud
2010 Lecture Notes in Computer Science  
The novelty resides both in the use of the Laplacian embedding to propagate must-link and cannotlink constraints, and in the segmentation algorithm which is based on a learn, align, transfer, and classify  ...  The method is composed of a constrained spectral clustering algorithm that is used to supervise the segmentation of a shape from a training data set, followed by a probabilistic label transfer algorithm  ...  Rather than modifying the K-means step of spectral clustering, we incorporate a constraint-propagation process directly into the L-embedding, thus fully exploiting the properties outlined in the previous  ... 
doi:10.1007/978-3-642-15555-0_54 fatcat:2tgpqmp25nfnffuhshtepkdefy

GP-S3Net: Graph-based Panoptic Sparse Semantic Segmentation Network [article]

Ryan Razani, Ran Cheng, Enxu Li, Ehsan Taghavi, Yuan Ren, Liu Bingbing
2021 arXiv   pre-print
We utilize the instance label to generate ground truth edge labels for each constructed graph in order to supervise the learning.  ...  Each cluster is treated as a node in the graph and its corresponding embedding is used as its node feature. Then a GCNN predicts whether edges exist between each cluster pair.  ...  The whole scene is usually segmented into small blocks, and the embeddings of each block are learned and used in the final clustering.  ... 
arXiv:2108.08401v1 fatcat:akhqf2euxjetzjccbtwt6t2csm
« Previous Showing results 1 — 15 out of 41,297 results