Filters








35,518 Hits in 5.3 sec

Scene Graph Prediction with Limited Labels [article]

Vincent S. Chen, Paroma Varma, Ranjay Krishna, Michael Bernstein, Christopher Re, Li Fei-Fei
2019 arXiv   pre-print
approach for training with limited labels.  ...  All scene graph models to date are limited to training on a small set of visual relationships that have thousands of training labels each.  ...  Training scene graph model: Finally, these probabilistic labels are used to train any scene graph prediction model.  ... 
arXiv:1904.11622v3 fatcat:3zoifzahgra5npffq3mtgkjhty

Scene Graph Prediction With Limited Labels

Ranjay Krishna, Vincent Chen, Paroma Varma, Michael Bernstein, Christopher Re, Li Fei-Fei
2019 2019 IEEE/CVF International Conference on Computer Vision (ICCV)  
approach for training with limited labels.  ...  All scene graph models to date are limited to training on a small set of visual relationships that have thousands of training labels each.  ...  To evaluate a scene graph model trained on our labels, we use three standard evaluation modes for scene graph prediction [31 ] : (i) scene graph detection (SGDET) which expects input images and predicts  ... 
doi:10.1109/iccv.2019.00267 pmid:32218709 pmcid:PMC7098690 fatcat:iyp65xdjurhbvog3fuxjhkc4zy

Learning and Reasoning with the Graph Structure Representation in Robotic Surgery [article]

Mobarakol Islam, Lalithkumar Seenivasan, Lim Chwee Ming, Hongliang Ren
2020 arXiv   pre-print
To obtain the graph scene label, we annotate the bounding box and the instrument-ROI interactions on the robotic scene segmentation challenge 2018 dataset with an experienced clinical expert in robotic  ...  For this purpose, we develop an approach to generate the scene graph and predict surgical interactions between instruments and surgical region of interest (ROI) during robot-assisted surgery.  ...  (c) Scene graph ground truth that represents the tissue-tool interaction in the scene. (d) Predicted scene graph based on our proposed model. (e) Predicted scene graph based on GPNN model [15] .  ... 
arXiv:2007.03357v3 fatcat:2ebz5psbnvhg7jqwsh4kelwv6m

Learning to Generate Scene Graph from Natural Language Supervision [article]

Yiwu Zhong, Jing Shi, Jianwei Yang, Chenliang Xu, Yin Li
2021 arXiv   pre-print
thus create "pseudo" labels for learning scene graph.  ...  Further, we design a Transformer-based model to predict these "pseudo" labels via a masked token prediction task.  ...  Acknowledgement: YZ and YL acknowledge the support provided by the UW-Madison OVCRGE with funding from WARF. JS and CX were supported by the National Science Foundation (NSF) under Grant RI:1813709.  ... 
arXiv:2109.02227v1 fatcat:s5nn6gezmjdethihx5traqhy6y

Multi-Label Remote Sensing Image Scene Classification by Combining a Convolutional Neural Network and a Graph Neural Network

Yansheng Li, Ruixian Chen, Yongjun Zhang, Mi Zhang, Ling Chen
2020 Remote Sensing  
Based on the trained CNN, one scene graph for each scene is further constructed, where nodes of the graph are represented by superpixel regions of the scene.  ...  As one of the fundamental tasks in remote sensing (RS) image understanding, multi-label remote sensing image scene classification (MLRSSC) is attracting increasing research interest.  ...  MLRSSC aims to predict multiple semantic labels to describe an RS image scene.  ... 
doi:10.3390/rs12234003 fatcat:nv3a55bhcre6xbd6lr2xbw4a7e

The Limited Multi-Label Projection Layer [article]

Brandon Amos, Vladlen Koltun, J. Zico Kolter
2019 arXiv   pre-print
The LML layer provides a probabilistic way of modeling multi-label predictions limited to having exactly k labels.  ...  We evaluate LML layers on top-k CIFAR-100 classification and scene graph generation.  ...  Multi-label predictions can arise from a task being truly multi-label, as in language and graph generation tasks, or by turning a single-label prediction task into a multi-label prediction task that predicts  ... 
arXiv:1906.08707v3 fatcat:hq6yx7scbvhihmouajay2jgrsu

Graph clustering-based crowd counting with very limited labelled samples

Huake Wang, Kaibing Zhang, Zebin Su, Jian Lu, Zenggang Xiong
2020 Electronics Letters  
In this Letter, the authors present a novel graph clustering-based method for crowd counting only using very limited labelled samples.  ...  These methods need to obtain the density map of crowd scenes by leveraging the annotations of individuals heads and convolving them with a Gaussian kernel.  ...  Conclusion: In this Letter, we present a novel graph clustering-based crowd counting with very limited labelled samples.  ... 
doi:10.1049/el.2020.0746 fatcat:fypqfekjrzdktbhcymsrprvqjm

Visual Distant Supervision for Scene Graph Generation [article]

Yuan Yao, Ao Zhang, Xu Han, Mengdi Li, Cornelius Weber, Zhiyuan Liu, Stefan Wermter, Maosong Sun
2021 arXiv   pre-print
However, scene graph models usually require supervised learning on large quantities of labeled data with intensive human annotation.  ...  In this work, we propose visual distant supervision, a novel paradigm of visual relation learning, which can train scene graph models without any human-labeled data.  ...  [56] reformulate scene graphs as bipartite graphs of objects and relations, and align the predicted graphs to their weakly supervised labels.  ... 
arXiv:2103.15365v2 fatcat:jjgqhd43qzerhcsq4z4i5gzmdm

Neural Motifs: Scene Graph Parsing with Global Context

Rowan Zellers, Mark Yatskar, Sam Thomson, Yejin Choi
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
Our analysis motivates a new baseline: given object detections, predict the most frequent relation between object pairs with the given labels, as seen in the training set.  ...  Our analysis shows that object labels are highly predictive of relation labels but not vice-versa.  ...  Acknowledgements We thank the anonymous reviewers along with Ali  ... 
doi:10.1109/cvpr.2018.00611 dblp:conf/cvpr/ZellersYTC18 fatcat:nvye4ywmyjdajpfei5zhqhn2lu

Neural Motifs: Scene Graph Parsing with Global Context [article]

Rowan Zellers, Mark Yatskar, Sam Thomson, Yejin Choi
2018 arXiv   pre-print
Our analysis motivates a new baseline: given object detections, predict the most frequent relation between object pairs with the given labels, as seen in the training set.  ...  Our analysis shows that object labels are highly predictive of relation labels but not vice-versa.  ...  Acknowledgements We thank the anonymous reviewers along with Ali  ... 
arXiv:1711.06640v2 fatcat:hvuai4lihzhwxhaog674btnqn4

Self-Supervised Real-to-Sim Scene Generation [article]

Aayush Prakash, Shoubhik Debnath, Jean-Francois Lafleche, Eric Cameracci, Gavriel State, Stan Birchfield, Marc T. Law
2021 arXiv   pre-print
We select scene graph (SG) generation as the downstream task, due to the limited availability of labeled datasets.  ...  To solve these challenges, we propose Sim2SG, a self-supervised automatic scene generation technique for matching the distribution of real data.  ...  The difficulty of hand-labeling scene graphs has limited the community to a small number of datasets [36, 29] .  ... 
arXiv:2011.14488v2 fatcat:gzp4oty5mjebtmq73fuugsdg7a

Graph-Based Embedding Smoothing Network for Few-Shot Scene Classification of Remote Sensing Images

Zhengwu Yuan, Wendong Huang, Chan Tang, Aixia Yang, Xiaobo Luo
2022 Remote Sensing  
Existing methods seek to take advantage of transfer knowledge or meta-knowledge to resolve the scene classification issue of remote sensing images with a handful of labeled samples while ignoring various  ...  labeled samples.  ...  Hence, our proposed method addresses the issues relevant to scene classification with limited labeled data by combining meta-learning.  ... 
doi:10.3390/rs14051161 fatcat:hvxzgzeg2bblvnpppdblgogd7y

Learning a Model for Inferring a Spatial Road Lane Network Graph using Self-Supervision [article]

Robin Karlsson, David Robert Wong, Simon Thompson, Kazuya Takeda
2021 arXiv   pre-print
A formal road lane network model is presented and proves that any structured road scene can be represented by a directed acyclic graph of at most depth three while retaining the notion of intersection  ...  This paper presents the first self-supervised learning method to train a model to infer a spatially grounded lane-level road network graph based on a dense segmented representation of the road scene generated  ...  ACKNOWLEDGMENT This research was supported by Program on Open Innovation Platform with Enterprises, Research Institute and Academia, Japan Science and Technology Agency (JST, OPERA, JPMJOP1612).  ... 
arXiv:2107.01784v1 fatcat:7lzq3dsxmfacro3nymf27jnzkm

Semantics-STGCNN: A Semantics-guided Spatial-Temporal Graph Convolutional Network for Multi-class Trajectory Prediction [article]

Ben A. Rainbow, Qianhui Men, Hubert P. H. Shum
2021 arXiv   pre-print
We embed the class labels of the surrounding objects into the label adjacency matrix (LAM), which is combined with the velocity-based adjacency matrix (VAM) comprised of the objects' velocity, thereby  ...  SAM effectively models semantic information with trainable parameters to automatically learn the embedded label features that will contribute to the fixed velocity-based trajectory.  ...  Furthermore, since our model is not limited to 2D video, it is also extendable to 3D trajectory with one more dimension included in the scene.  ... 
arXiv:2108.04740v1 fatcat:ueqfb7vc5fd7jgehep3wttqbzy

Differentiable Scene Graphs [article]

Moshiko Raboh, Roei Herzig, Gal Chechik, Jonathan Berant, Amir Globerson
2020 arXiv   pre-print
Scene graphs provide a natural representation for reasoning tasks, by assigning labels to both entities (nodes) and relations (edges).  ...  Here we propose Differentiable Scene Graphs (DSGs), an image representation that is amenable to differentiable end-to-end optimization, and requires supervision only from the downstream tasks.  ...  Given these limitations, it is an open question how to make scene graphs useful for visual reasoning applications.  ... 
arXiv:1902.10200v5 fatcat:sdm4ctswj5hzpdhvluya67vpei
« Previous Showing results 1 — 15 out of 35,518 results