2,459 Hits in 3.8 sec

Unsupervised Semantic Segmentation by Distilling Feature Correspondences [article]

Mark Hamilton, Zhoutong Zhang, Bharath Hariharan, Noah Snavely, William T. Freeman
2022 arXiv   pre-print
Unsupervised semantic segmentation aims to discover and localize semantically meaningful categories within image corpora without any form of annotation.  ...  This observation motivates us to design STEGO (Self-supervised Transformer with Energy-based Graph Optimization), a novel framework that distills unsupervised features into high-quality discrete semantic  ...  This material is based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant No. 2021323067.  ... 
arXiv:2203.08414v1 fatcat:eg6waeo7obelxgporlx4a2t2wi

Drive Segment: Unsupervised Semantic Segmentation of Urban Scenes via Cross-modal Distillation [article]

Antonin Vobecky, David Hurych, Oriane Siméoni, Spyros Gidaris, Andrei Bursuc, Patrick Pérez, Josef Sivic
2022 arXiv   pre-print
First, we propose a novel method for cross-modal unsupervised learning of semantic image segmentation by leveraging synchronized LiDAR and image data.  ...  Finally, we develop a cross-modal distillation approach that leverages image data partially annotated with the resulting pseudo-classes to train a transformer-based model for image semantic segmentation  ...  Antonin Vobecky was partly supported by the CTU Student Grant Agency (reg. no. SGS21184OHK33T37).  ... 
arXiv:2203.11160v1 fatcat:6tpiuvzqxnexja2wym66d6uirq

Domain Adaptive Knowledge Distillation for Driving Scene Semantic Segmentation [article]

Divya Kothandaraman, Athira Nambiar, Anurag Mittal
2020 arXiv   pre-print
We term this as "Domain Adaptive Knowledge Distillation" and address the same in the context of unsupervised domain-adaptive semantic segmentation by proposing a multi-level distillation strategy to effectively  ...  images, where the problem is completely unsupervised.  ...  semantic segmentation.  ... 
arXiv:2011.08007v2 fatcat:phzdkoce6nhmfiv5fkjhhwmp44

Segmentation in Style: Unsupervised Semantic Image Segmentation with Stylegan and CLIP [article]

Daniil Pakhomov, Sanchit Hira, Narayani Wagle, Kemar E. Green, Nassir Navab
2021 arXiv   pre-print
Once classes are discovered, a synthetic dataset with generated images and corresponding segmentation masks can be created.  ...  Additionally, by using CLIP we are able to use prompts defined in a natural language to discover some desired semantic classes.  ...  Inspired by recent work in knowledge distillation for image manipulation with Style-GAN [19] we generate a synthetic dataset with images and corresponding segmentation masks and train a simple segmentation  ... 
arXiv:2107.12518v2 fatcat:xkj4pjigrnegxeyv34xrj4oaeq

BoMuDANet: Unsupervised Adaptation for Visual Scene Understanding in Unstructured Driving Environments [article]

Divya Kothandaraman, Rohan Chandra, Dinesh Manocha
2021 arXiv   pre-print
We describe a new semantic segmentation technique based on unsupervised domain adaptation (DA), that can identify the class or category of each region in RGB images or videos.  ...  A key aspect of our approach is that it can also identify objects that are encountered by the model for the fist time during the testing phase.  ...  We present the first method for unsupervised multisource boundless domain adaptive semantic segmentation (See Figure 2 ).  ... 
arXiv:2010.03523v3 fatcat:72sr7avhhjb7niwkheizlc2eeu

Discovering Object Masks with Transformers for Unsupervised Semantic Segmentation [article]

Wouter Van Gansbeke, Simon Vandenhende, Luc Van Gool
2022 arXiv   pre-print
By combining these components, we can considerably outperform previous works for unsupervised semantic segmentation on PASCAL (+11% mIoU) and COCO (+4% mask AP50).  ...  The task of unsupervised semantic segmentation aims to cluster pixels into semantically meaningful groups.  ...  The authors thankfully acknowledge support by Toyota Motor Europe (TME) via the TRACE project.  ... 
arXiv:2206.06363v1 fatcat:o5op2tqhvnb7bnro7vxbwvvaga

Multi-Head Distillation for Continual Unsupervised Domain Adaptation in Semantic Segmentation [article]

Antoine Saporta and Arthur Douillard and Tuan-Hung Vu and Patrick Pérez and Matthieu Cord
2022 arXiv   pre-print
Unsupervised Domain Adaptation (UDA) is a transfer learning task which aims at training on an unlabeled target domain by leveraging a labeled source domain.  ...  MuHDi performs distillation at multiple levels from the previous model as well as an auxiliary target-specialist segmentation head.  ...  Multi-Head Distillation Framework This section presents our approach to continual UDA for semantic segmentation: MuHDi, for Multi-Head Distillation.  ... 
arXiv:2204.11667v1 fatcat:m3hjm2q6h5euzbc4jiybjwprnu

Disentangled Latent Transformer for Interpretable Monocular Height Estimation [article]

Zhitong Xiong, Sining Chen, Yilei Shi, Xiao Xiang Zhu
2022 arXiv   pre-print
Furthermore, a novel unsupervised semantic segmentation task based on height estimation is first introduced in this work.  ...  Additionally, we also construct a new dataset for joint semantic segmentation and height estimation. Our work provides novel insights for both understanding and designing MHE models.  ...  The results in Table 5 show that DLT can outperform existing unsupervised semantic segmentation methods clearly by a large margin.  ... 
arXiv:2201.06357v2 fatcat:g3zpj7pr2fg2jhkau6t52wvaqy

Multi-Target Domain Adaptation with Collaborative Consistency Learning [article]

Takashi Isobe, Xu Jia, Shuaijun Chen, Jianzhong He, Yongjie Shi, Jianzhuang Liu, Huchuan Lu, Shengjin Wang
2021 arXiv   pre-print
Recently unsupervised domain adaptation for the semantic segmentation task has become more and more popular due to high-cost of pixel-level annotation on real-world images.  ...  These expert models are further improved by adding the regularization of making the consistent pixel-wise prediction for each sample with the same structured context.  ...  Introduction Semantic segmentation aims at interpreting an image by assigning each pixel to a semantic class [33, 6, 7, 55, 63] .  ... 
arXiv:2106.03418v1 fatcat:l63vu5efafaatpr6fktudjwe7y

Deep Spectral Methods: A Surprisingly Strong Baseline for Unsupervised Semantic Segmentation and Localization [article]

Luke Melas-Kyriazi and Christian Rupprecht and Iro Laina and Andrea Vedaldi
2022 arXiv   pre-print
Furthermore, by clustering the features associated with these segments across a dataset, we can obtain well-delineated, nameable regions, i.e. semantic segmentations.  ...  Unsupervised localization and segmentation are long-standing computer vision challenges that involve decomposing an image into semantically-meaningful segments without any labeled data.  ...  unsupervised object localization, object segmentation, semantic segmentation and image matting.  ... 
arXiv:2205.07839v1 fatcat:6iyz22evgzak5h2apmxbywowuq

Invariant Information Clustering for Unsupervised Image Classification and Segmentation [article]

Xu Ji, João F. Henriques, Andrea Vedaldi
2019 arXiv   pre-print
The model discovers clusters that accurately match semantic classes, achieving state-of-the-art results in eight unsupervised clustering benchmarks spanning image classification and segmentation.  ...  These include STL10, an unsupervised variant of ImageNet, and CIFAR10, where we significantly beat the accuracy of our closest competitors by 8 and 9.5 absolute percentage points respectively.  ...  The clusters found by IIC are highly discriminative (fig. although note some failure cases; as IIC distills purely visual correspondences within images, it can be confused by instances that combine classes  ... 
arXiv:1807.06653v3 fatcat:f4u7gexfazcrjiwp7ascx5xylu

A Survey on Label-efficient Deep Segmentation: Bridging the Gap between Weak Supervision and Dense Prediction [article]

Wei Shen, Zelin Peng, Xuehui Wang, Huayu Wang, Jiazhong Cen, Dongsheng Jiang, Lingxi Xie, Xiaokang Yang, Qi Tian
2022 arXiv   pre-print
and noisy supervision) and supplemented by the types of segmentation problems (including semantic segmentation, instance segmentation and panoptic segmentation).  ...  To this end, we first develop a taxonomy to organize these methods according to the supervision provided by different types of weak labels (including no supervision, coarse supervision, incomplete supervision  ...  ACKNOWLEDGMENTS This work was supported by NSFC 62176159, Natural Science Foundation of Shanghai 21ZR1432200 and Shanghai Municipal Science and Technology Major Project 2021SHZDZX0102.  ... 
arXiv:2207.01223v1 fatcat:i7rgpxrfkrdbfm4effjdcjjr24

Cross-Domain Correlation Distillation for Unsupervised Domain Adaptation in Nighttime Semantic Segmentation [article]

Huan Gao, Jichang Guo, Guoli Wang, Qian Zhang
2022 arXiv   pre-print
The performance of nighttime semantic segmentation is restricted by the poor illumination and a lack of pixel-wise annotation, which severely limit its application in autonomous driving.  ...  Extensive experiments on Dark Zurich and ACDC demonstrate that CCDistill achieves the state-of-the-art performance for nighttime semantic segmentation.  ...  Conclusions In this paper, we propose an unsupervised domain adaptation framework via the invariance of cross-domain difference for nighttime semantic segmentation.  ... 
arXiv:2205.00858v1 fatcat:m65upvhgfjdbdd6bnl5ftnzluu

Detecting and Learning the Unknown in Semantic Segmentation [article]

Robin Chan, Svenja Uhlemeyer, Matthias Rottmann, Hanno Gottschalk
2022 arXiv   pre-print
Next, we review research in detecting semantically unknown objects in semantic segmentation.  ...  Semantic segmentation is a crucial component for perception in automated driving.  ...  Acknowledgments The research leading to these results is funded by the German Federal Ministry for Economic Affairs and Climate Action within the projects "KI Absicherung -Safe AI for Automated Driving  ... 
arXiv:2202.08700v1 fatcat:6e6vvviq2zed5idpuv3iuqre7i

Bootstrapping Semi-supervised Medical Image Segmentation with Anatomical-aware Contrastive Distillation [article]

Chenyu You, Weicheng Dai, Lawrence Staib, James S. Duncan
2022 arXiv   pre-print
In this work, we present ACTION, an Anatomical-aware ConTrastive dIstillatiON framework, for semi-supervised medical image segmentation.  ...  We also capture more semantically similar features from the randomly chosen negative set compared to the positives to enforce the diversity of the sampled data.  ...  We use the supervised segmentation loss on labeled data, unsupervised cross-entropy loss (on pseudo-labels generated by a confidence threshold θ s ), and L anco in Equation 4 on unlabeled data.  ... 
arXiv:2206.02307v1 fatcat:6uygotfvz5anljiz5gwmaz54pq
« Previous Showing results 1 — 15 out of 2,459 results