Filters








19,932 Hits in 5.4 sec

Transfer Learning Improves Supervised Image Segmentation Across Imaging Protocols

Annegreet van Opbroek, M. Arfan Ikram, Meike W. Vernooij, Marleen de Bruijne
2015 IEEE Transactions on Medical Imaging  
Transfer learning improves supervised image segmentation across imaging protocols van Opbroek, Annegreet; Ikram, M.  ...  Transfer learning improves supervised image segmentation across imaging protocols. IEEE transactions on medical imaging, 34(5), 1018-1030.  ...  The purpose of our study was to investigate whether transfer-learning techniques can improve upon regular supervised segmentation of images obtained with different scan protocols.  ... 
doi:10.1109/tmi.2014.2366792 pmid:25376036 fatcat:nhngb27plrd3tbgte337p6c2ay

Using Rule-Based Labels for Weak Supervised Learning: A ChemNet for Transferable Chemical Property Prediction [article]

Garrett B. Goh, Charles Siegel, Abhinav Vishnu, Nathan O. Hodas
2018 arXiv   pre-print
DNN models that were trained using conventional supervised learning.  ...  In this work, we develop an approach of using rule-based knowledge for training ChemNet, a transferable and generalizable deep neural network for chemical property prediction that learns in a weak-supervised  ...  Nathan Baker for helpful discussions. is work is supported by the following PNNL LDRD programs: Pauling Postdoctoral Fellowship and Deep Learning for Scienti c Discovery Agile Investment.  ... 
arXiv:1712.02734v2 fatcat:itrjobfzkzexnlw5nqwxjqmzk4

Efficient Visual Pretraining with Contrastive Detection [article]

Olivier J. Hénaff, Skanda Koppula, Jean-Baptiste Alayrac, Aaron van den Oord, Oriol Vinyals, João Carreira
2021 arXiv   pre-print
Finally, our objective seamlessly handles pretraining on more complex images such as those in COCO, closing the gap with supervised transfer learning from COCO to PASCAL.  ...  Self-supervised pretraining has been shown to yield powerful representations for transfer learning.  ...  Results: larger models In Table 2 we compare to prior works on self-supervised learning which transfer to COCO.  ... 
arXiv:2103.10957v2 fatcat:xdpkl5tr6ff3xf2tm5bj42sd6u

CYBORGS: Contrastively Bootstrapping Object Representations by Grounding in Segmentation [article]

Renhao Wang, Hang Zhao, Yang Gao
2022 arXiv   pre-print
Experiments show our representations transfer robustly to downstream tasks in classification, detection and segmentation.  ...  Many recent approaches in contrastive learning have worked to close the gap between pretraining on iconic images like ImageNet and pretraining on complex scenes like COCO.  ...  Main Results: Representation Learning We follow standard downstream transfer-based protocols to evaluate the strength of representations learned by cyborgs.  ... 
arXiv:2203.09343v1 fatcat:w55qgtxkcfb5hdnzlljmfbuydy

Learning Transferrable Knowledge for Semantic Segmentation with Deep Convolutional Neural Network [article]

Seunghoon Hong, Junhyuk Oh, Bohyung Han, Honglak Lee
2015 arXiv   pre-print
Contrary to existing weakly-supervised approaches, our algorithm exploits auxiliary segmentation annotations available for different categories to guide segmentations on images with only image-level class  ...  To make the segmentation knowledge transferrable across categories, we design a decoupled encoder-decoder architecture with attention model.  ...  is appropriate to transfer the segmentation knowledge across categories. • The proposed algorithm achieved substantial performance improvement over existing weakly-supervised approaches with segmentation  ... 
arXiv:1512.07928v1 fatcat:p6kdgj7gbvdrrgxzrdhss275k4

Advancing Medical Imaging Informatics by Deep Learning-Based Domain Adaptation

Anirudh Choudhary, Li Tong, Yuanda Zhu, May D. Wang
2020 IMIA Yearbook of Medical Informatics  
DA is a type of transfer learning (TL) that can improve the performance of models when applied to multiple different datasets.  ...  , image modality, and learning scenarios.  ...  ., [77] have applied a domain discriminator to MR images from different scanners and imaging protocols to improve the brain lesion segmentation performance.  ... 
doi:10.1055/s-0040-1702009 pmid:32823306 fatcat:gtlhoh6m3fh4hcumfzdlpdohr4

Learning Transferrable Knowledge for Semantic Segmentation with Deep Convolutional Neural Network

Seunghoon Hong, Junhyuk Oh, Honglak Lee, Bohyung Han
2016 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)  
Contrary to existing weakly-supervised approaches, our algorithm exploits auxiliary segmentation annotations available for different categories to guide segmentations on images with only image-level class  ...  To make segmentation knowledge transferrable across categories, we design a decoupled encoder-decoder architecture with attention model.  ...  is appropriate to transfer the segmentation knowledge across categories. • The proposed algorithm achieves substantial performance improvement over existing weakly-supervised approaches by exploiting  ... 
doi:10.1109/cvpr.2016.349 dblp:conf/cvpr/HongOLH16 fatcat:obemojsyvvfflodeaxd3ptfyse

Unsupervised Semantic Segmentation by Contrasting Object Mask Proposals [article]

Wouter Van Gansbeke, Simon Vandenhende, Stamatios Georgoulis, Luc Van Gool
2021 arXiv   pre-print
Being able to learn dense semantic representations of images without supervision is an important problem in computer vision.  ...  Second, our representations can improve over strong baselines when transferred to new datasets, e.g. COCO and DAVIS. The code is available.  ...  Interestingly, our representations transfer well across various datasets.  ... 
arXiv:2102.06191v3 fatcat:6kbodv6wbjblzorfimrs4hz5v4

Learning from 2D: Contrastive Pixel-to-Point Knowledge Transfer for 3D Pretraining [article]

Yueh-Cheng Liu, Yu-Kai Huang, Hung-Yueh Chiang, Hung-Ting Su, Zhe-Yu Liu, Chin-Tang Chen, Ching-Yu Tseng, Winston H. Hsu
2021 arXiv   pre-print
In this paper, we present a novel 3D pretraining method by leveraging 2D networks learned from rich 2D datasets.  ...  Our intensive experiments show that the 3D models pretrained with 2D knowledge boost the performances of 3D networks across various real-world 3D downstream tasks.  ...  Following the pretrain-finetune protocol in [64] , we show that PPKT consistently boosts overall downstream performance across multiple real-world 3D tasks and datasets.  ... 
arXiv:2104.04687v3 fatcat:zfmpvxsv6vevlfcualoorajmkm

Object discovery and representation networks [article]

Olivier J. Hénaff, Skanda Koppula, Evan Shelhamer, Daniel Zoran, Andrew Jaegle, Andrew Zisserman, João Carreira, Relja Arandjelović
2022 arXiv   pre-print
Instead, we propose a self-supervised learning paradigm that discovers this image structure by itself.  ...  The resulting learning paradigm is simpler, less brittle, and more general, and achieves state-of-the-art transfer learning results for object detection and instance segmentation on COCO, and semantic  ...  Contrastive learning departed from this tradition by radically simplifying the self-supervised protocol, in that the pretext task is specified by the data itself: representations must learn to distinguish  ... 
arXiv:2203.08777v2 fatcat:vkem4cnhpnhf3bqebzusk4o45m

Self-Supervised Learning from Unlabeled Fundus Photographs Improves Segmentation of the Retina [article]

Jan Kukačka, Anja Zenz, Marcel Kollovieh, Dominik Jüstel, Vasilis Ntziachristos
2021 arXiv   pre-print
To overcome these limitations, we utilized contrastive self-supervised learning to exploit the large variety of unlabeled fundus images in the publicly available EyePACS dataset.  ...  Automated segmentation of fundus photographs would improve the quality, capacity, and cost-effectiveness of eye care screening programs.  ...  We repeated the experiments several times (N=4 for image segmentation, N=12 for domain transfer) and paired the results from matching training/validation splits.  ... 
arXiv:2108.02798v1 fatcat:fcmedcintnechhbxybt44lhfhy

What makes instance discrimination good for transfer learning? [article]

Nanxuan Zhao and Zhirong Wu and Rynson W.H. Lau and Stephen Lin
2021 arXiv   pre-print
It comes as a surprise that image annotations would be better left unused for transfer learning.  ...  In this work, we investigate the following problems: What makes instance discrimination pretraining good for transfer learning? What knowledge is actually learned and transferred from these models?  ...  This is in contrast to traditional supervised learning, where ImageNet performance is improved and its transfer performance is compromised.  ... 
arXiv:2006.06606v2 fatcat:gyleg63lbzfqpbkb2b3aryz63u

Domain Adaptive Relational Reasoning for 3D Multi-Organ Segmentation [article]

Shuhao Fu, Yongyi Lu, Yan Wang, Yuyin Zhou, Wei Shen, Elliot Fishman, Alan Yuille
2020 arXiv   pre-print
as a latent variable to transfer the knowledge shared across multiple domains.  ...  To guarantee the transferability of the learned spatial relationship to multiple domains, we additionally introduce two schemes: 1) Employing a super-resolution network also jointly trained with the segmentation  ...  Such relational configuration is deemed as weak cues for segmentation task, which is easier to learn, and thus better in transfer [28] .  ... 
arXiv:2005.09120v2 fatcat:uaoaehddgffhlh3btpresyrb5y

Exploring Set Similarity for Dense Self-supervised Representation Learning [article]

Zhaoqing Wang, Qiang Li, Guoxin Zhang, Pengfei Wan, Wen Zheng, Nannan Wang, Mingming Gong, Tongliang Liu
2022 arXiv   pre-print
Meanwhile, these attentional features can keep the coherence of the same image across different views to alleviate semantic inconsistency.  ...  We generalize pixel-wise similarity learning to set-wise one to improve the robustness because sets contain more semantic and structure information.  ...  Conclusion In this paper, we propose a simple but effective dense self-supervised representation learning framework, SetSim, by exploring set similarity across views to improve the robustness.  ... 
arXiv:2107.08712v2 fatcat:jqtpzpseird2hbldlpjj4x7txa

VirTex: Learning Visual Representations from Textual Annotations [article]

Karan Desai, Justin Johnson
2021 arXiv   pre-print
We train convolutional networks from scratch on COCO Captions, and transfer them to downstream recognition tasks including image classification, object detection, and instance segmentation.  ...  On all tasks, VirTex yields features that match or exceed those learned on ImageNet -- supervised or unsupervised -- despite using up to ten times fewer images.  ...  model zoo; Georgia Gkioxari for suggesting the Instance Segmentation pretraining task ablation; and Stefan Lee for suggestions on figure aesthetics.  ... 
arXiv:2006.06666v3 fatcat:ifck6jbayvc4hcrznk6icqghga
« Previous Showing results 1 — 15 out of 19,932 results