IP Library › Granted Patent US 12,675,983
Granted Patent B2
US 12,675,983 · App. 18/103,119 · Granted Jul 7, 2026

Systems and methods for semantic image segmentation model learning new object classes

Inventors: Subhankar Roy (Kolkata, IN); Riccardo Volpi (Grenoble, FR); Diane Larlus (La Tronche, FR); Gabriela Csurka Khedari (Crolles, FR)
Assignee: NAVER CORPORATION
G06V10/7753G05D1/0246G06V10/26G06V10/761G06V10/764G06V20/70G06V10/776
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,675,983
App. No.
18/103,119
Granted
Jul 7, 2026
Kind
B2
Abstract

A semantic image segmentation (SIS) system includes: a semantic segmentation module trained to segment objects belonging to predetermined classes in input images using training images; and a learning module configured to selectively update at least one parameter of each of a localizer module, an encoder module, and a decoder module of the semantic segmentation module to identify objects having a new class that is not one of the predetermined classes: based on an image level class for a learning image including an object having the new class that is not one of the predetermined classes; and without a pixel-level annotation for the learning image.

Claims (60)

1 . A semantic image segmentation (SIS) system, comprising:

a semantic segmentation module trained to segment objects belonging to predetermined classes in input images using training images;

a learning module configured to selectively update at least one parameter of each of a localizer module, an encoder module, and a decoder module of the semantic segmentation module to identify objects having a new class that is not one of the predetermined classes:

based on an image level class for a learning image including an object having the new class that is not one of the predetermined classes; and

without a pixel-level annotation for the learning image; and

a semantic map module configured to determine semantic similarity measures between the new class and the predetermined classes, respectively,

wherein the learning module is configured to selectively update at least one parameter of the localizer module of the semantic segmentation module based on the at least one of the determined semantic similarity measures.

2 . The SIS system of claim 1 wherein

the semantic map module is configured to generate semantic similarity maps based on the determined semantic similarity measure between the new class and the predetermined classes, respectively,

wherein the learning module is configured to selectively update at least one parameter of the localizer module of the semantic segmentation module based on the semantic similarity maps.

3 . The SIS system of claim 2 further comprising a semantic loss module configured to determine a loss based on the semantic similarity maps,

wherein the learning module is configured to selectively update at least one parameter of the localizer module of the semantic segmentation module based on the loss.

4 . The SIS system of claim 3 wherein the learning module is configured to selectively update at least one parameter of the localizer module of the semantic segmentation module based on minimizing the loss.

5 . The SIS system of claim 1 wherein the semantic segmentation module is trained to segment objects belonging to the predetermined classes in input images using training images and pixel-level annotations for training images including objects having the predetermined classes, the pixel-level annotations including pixels defining boundaries of the objects in the training images.

6 . The SIS system of claim 1 wherein the learning module is configured to update a classifier layer of the decoder module of the semantic segmentation module.

7 . The SIS system of claim 1 further comprising a segmentation loss module configured to determine a loss,

wherein the learning module is configured to selectively update the at least one parameter of at least one of (a) the encoder module of the semantic segmentation module, (b) the decoder module of the semantic segmentation module, and (c) the localizer module based on the loss.

8 . The SIS system of claim 7 wherein the learning module is configured to selectively update the at least one parameter of at least one of (a) the encoder module of the semantic segmentation module, (b) the decoder module of the semantic segmentation module, and (c) the localizer module based on minimizing the loss.

9 . The SIS system of claim 1 wherein the one of the predetermined classes is a background class corresponding to background behind objects.

10 . A robot comprising:

a camera;

the semantic image segmentation system of claim 1 ; and

a control module configured to actuate an actuator of the robot based on an object segmented from an image from the camera by the semantic segmentation module.

11 . The SIS system of claim 1 wherein the new class includes a word descriptive of the object in the learning image.

12 . The SIS system of claim 1 wherein the learning module is further configured to selectively update at least one parameter of each of the localizer module, the encoder module, and the decoder module of the semantic segmentation module to identify objects having a second new class that is not the new class and not one of the predetermined classes:

based on a second image level class for a second learning image including a second object having the second new class that is not one of the predetermined classes and not the new class; and

without a pixel-level annotation for the second learning image.

13 . A semantic image segmentation (SIS) system, comprising:

a semantic segmentation module including a semantic segmentation model that receives a first training using a first set of training images with pixel-level annotations labeled with one or more predetermined classes, the semantic segmentation module configured to segment objects in an input image using the semantic segmentation model into one or more of the predetermined classes; and

a learning module configured to selectively update at least one parameter of the semantic segmentation model using a second set of training images with image-level annotations and without pixel-level annotations,

the semantic segmentation model receiving a second training from the learning module to identify objects of a first new class that is not one of the predetermined classes using semantic similarity measures determined between labels identifying the first new class and labels identifying the predetermined classes, respectively,

wherein the semantic segmentation module is configured to segment objects in the input image into one or more of the predetermined classes and the first new class once the segmentation model is second trained by the learning module for the first new class.

14 . The SIS system of claim 13 wherein:

the learning module is configured to selectively update at least one parameter of the semantic segmentation model using a third set of training images with image-level annotations without pixel-level annotations;

the semantic segmentation model receives a third training from the learning module to identify objects of a second new class that is not one of the predetermined classes or the first new class using semantic similarity measures determined between labels identifying the second new class and labels identifying the predetermined classes and the first new class, respectively; and

the semantic segmentation module segments objects in the input image into one or more of the predetermined classes, the first new class, and the second new class once the segmentation model is trained by the learning module for the second new class.

15 . A semantic image segmentation (SIS) system, comprising:

memory storing a semantic segmentation model trained using (a) a first set of training images with pixel-level annotations labeled for a first set of classes before being trained with (b) a second set of training images and image-level annotations without pixel-level annotations labeled with a first new class not in the first set of classes using semantic similarity measures determined between labels identifying the first new class and labels identifying the first set of classes, respectively; and

a semantic segmentation module configured to, after the training, segment objects in an input image into the first set of classes and the first new class using the semantic segmentation model.

16 . A semantic image segmentation (SIS) method, comprising:

obtaining a semantic segmentation module trained to segment objects belonging to predetermined classes in input images using training images;

selectively updating at least one parameter of each of a localizer module, an encoder module, and a decoder module of the semantic segmentation module to identify objects having a new class that is not one of the predetermined classes:

based on an image level class for a learning image including an object having the new class that is not one of the predetermined classes; and

without a pixel-level annotation for the learning image; and

determining semantic similarity measures between the new class and the predetermined classes, respectively,

wherein the selectively updating includes selectively updating at least one parameter of the localizer module of the semantic segmentation module based on the determined semantic similarity measures.

17 . The SIS method of claim 16 further comprising:

generating semantic similarity maps based on the determined semantic similarity measure between the new class and the predetermined classes, respectively,

wherein the selectively updating includes selectively updating at least one parameter of the localizer module of the semantic segmentation module based on the semantic similarity maps.

18 . The SIS method of claim 17 further comprising determining a loss based on the semantic similarity maps,

wherein the selectively updating includes selectively updating at least one parameter of the localizer module of the semantic segmentation module based on the loss.

19 . The SIS method of claim 18 wherein the selectively updating includes selectively updating at least one parameter of the localizer module of the semantic segmentation module based on minimizing the loss.

20 . The SIS method of claim 16 wherein the semantic segmentation module is trained to segment objects belonging to the predetermined classes in input images using training images and pixel-level annotations for training images including objects having the predetermined classes, the pixel-level annotations including pixels defining boundaries of the objects in the training images.

21 . The SIS method of claim 16 wherein the selectively updating includes updating a classifier layer of the decoder module of the semantic segmentation module.

22 . A semantic image segmentation (SIS) system, comprising:

a semantic segmentation module trained to segment objects belonging to predetermined classes in input images using training images;

a learning module configured to selectively update at least one parameter of each of a localizer module, an encoder module, and a decoder module of the semantic segmentation module to identify objects having a new class that is not one of the predetermined classes:

based on an image level class for a learning image including an object having the new class that is not one of the predetermined classes; and

without a pixel-level annotation for the learning image,

wherein the learning module is configured to update a classifier layer of the decoder module of the semantic segmentation module.

Assignments (2)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Sep 5, 2024
From: NAVER LABS CORPORATION
To: NAVER CORPORATION
Reel/Frame 068820/0495 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jan 30, 2023
From: ROY, SUBHANKAR; VOLPI, RICCARDO; LARLUS, DIANE; CSURKA KHEDARI, GABRIELA
To: CORPORATION, NAVER; NAVER LABS CORPORATION
Reel/Frame 062535/0076 →
Continuity (1)
Related Publication 20240257504A1 · Aug 1, 2024
References Cited (90)
US 11809523B2 · Kastaniotis et al. · 2023 [cited by examiner]
US 12286115B2 · Pham et al. · 2025 [cited by examiner]
US 20170330059A1 · Novotny et al. · 2017 [cited by examiner]
US 20220076117A1 · Amon et al. · 2022 [cited by examiner]
CN 117009560A · 2023 [cited by examiner]
“Incremental Learning Techniques for Semantic Segmentation”, by Michieli et al., published in 2019 in 2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW) (Year: 2019). [cited by examiner]
“Incremental Learning in Semantic Segmentation from Image Labels”, by Cermelli et al., published in Jun. 2022 in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), New Orleans, LA, USA, 2022, pp… [cited by examiner]
Ahn, Jiwoon, and Suha Kwak. “Learning Pixel-Level Semantic Affinity with Image-Level Supervision for Weakly Supervised Semantic Segmentation.” In [cited by applicant]
Araslanov, Nikita, and Stefan Roth. “Single-Stage Semantic Segmentation From Image Labels.” In [cited by applicant]
Baek, Donghyeon, Youngmin Oh, and Bumsub Ham. “Exploiting a Joint Embedding Space for Generalized Zero-Shot Semantic Segmentation.” In [cited by applicant]
Bearman, Amy, Olga Russakovsky, Vittorio Ferrari, and Li Fei-Fei. “What's the Point: Semantic Segmentation with Point Supervision.” In [cited by applicant]
Borenstein, Eran, and Shimon Ullman. “Learning to Segment.” In [cited by applicant]
Bucher, Maxime, Tuan-Hung Vu, Matthieu Cord, and Patrick Pérez. “Zero-Shot Semantic Segmentation,” NeurIPS, 2019. [cited by applicant]
Cermelli, Fabio. “Prototype-Based Incremental Few-Shot Semantic Segmentation,” BMVC, 2021. [cited by applicant]
Cermelli, Fabio, Dario Fontanel, Antonio Tavera, Marco Ciccone, and Barbara Caputo. “Incremental Learning in Semantic Segmentation from Image Labels.” In [cited by applicant]
Cermelli, Fabio, Massimiliano Mancini, Samuel Rota Bulo, Elisa Ricci, and Barbara Caputo. “Modeling the Background for Incremental Learning in Semantic Segmentation.” In [cited by applicant]
Cha, Sungmin, Beomyoung Kim, YoungJoon Yoo, and Taesup Moon. “SSUL: Semantic Segmentation with Unknown Label for Exemplar-Based Class-Incremental Learning,” NeurIPS 2021. [cited by applicant]
Cordts, Marius, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus Enzweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. “The Cityscapes Dataset for Semantic Urban Scene Understanding.” In [cited by applicant]
Dai, Jifeng, Kaiming He, and Jian Sun. “BoxSup: Exploiting Bounding Boxes to Supervise Convolutional Networks for Semantic Segmentation.” In [cited by applicant]
Devlin, Jacob, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. “BERT: Pre-Training of Deep Bidirectional Transformers for Language Understanding,” Proceedings of NAACL-HLT, 2019. [cited by applicant]
Ding, Henghui, Chang Liu, Suchen Wang, and Xudong Jiang. “Vision-Language Transformer and Query Generation for Referring Segmentation.” In [cited by applicant]
Douillard, Arthur, Yifu Chen, Arnaud Dapogny, and Matthieu Cord. “PLOP: Learning without Forgetting for Continual Semantic Segmentation.” In [cited by applicant]
Gidaris, Spyros, and Nikos Komodakis. “Dynamic Few-Shot Visual Learning Without Forgetting.” In [cited by applicant]
Gomez, Lluis, Yash Patel, Marcal Rusinol, Dimosthenis Karatzas, and C. V. Jawahar. “Self-Supervised Learning of Visual Features through Embedding Images into Text Topic Spaces.” In [cited by applicant]
He, Kaiming, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. “Deep Residual Learning for Image Recognition.” In [cited by applicant]
Hu, Ronghang, Marcus Rohrbach, and Trevor Darrell. “Segmentation from Natural Language Expressions.” arXiv, Mar. 20, 2016. [cited by applicant]
Huang, Zilong, Xinggang Wang, Jiasi Wang, Wenyu Liu, and Jingdong Wang. “Weakly-Supervised Semantic Segmentation Network with Deep Seeded Region Growing.” In [cited by applicant]
Ioffe, Sergey. “Batch Renormalization: Towards Reducing Minibatch Dependence in Batch-Normalized Models,” NIPS, 2017. [cited by applicant]
Ji, Zongliang. “Weakly Supervised Semantic Segmentation: From Box to Tag and Back,” BMVC, 2021. [cited by applicant]
Jia, Chao, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V Le, Yunhsuan Sung, Zhen Li, and Tom Duerig. “Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision,” ICM… [cited by applicant]
Kolesnikov, Alexander, and Christoph H. Lampert. “Seed, Expand and Constrain: Three Principles for Weakly-Supervised Image Segmentation.” In [cited by applicant]
Kulharia, Viveka, Siddhartha Chandra, Amit Agrawal, Philip Torr, and Ambrish Tyagi. “Box2Seg: Attention Weighted Loss and Discriminative Feature Learning for Weakly Supervised Segmentation.” In [cited by applicant]
Li, Boyi, Vladlen Koltun, Kilian Q Weinberger, René Ranftl, and Serge Belongie. “Language-Driven Semantic Segmentation,” ICLR, 2022. [cited by applicant]
Li, Peike, Yunchao Wei, and Yi Yang. “Consistent Structural Relation Learning for Zero-Shot Segmentation,” NeurIPS 2020. [cited by applicant]
Li, Zhizhong, and Derek Hoiem. “Learning without Forgetting.” arXiv. [cited by applicant]
Lin, Di, Jifeng Dai, Jiaya Jia, Kaiming He, and Jian Sun. “ScribbleSup: Scribble-Supervised Convolutional Networks for Semantic Segmentation.” In [cited by applicant]
Lin, Tsung-Yi, Maire, Michael, Belongie, Serge, Hays, James, Perona, Pietro, Ramanan, Deva, Dollar, Piotr, and Zitnick, C. Lawrence. “Microsoft COCO: Common Objects in Context.” arXiv:1405.0312. Feb. 21, 2015. [cited by applicant]
Liu, Chenxi, Zhe Lin, Xiaohui Shen, Jimei Yang, Xin Lu, and Alan Yuille. “Recurrent Multimodal Interaction for Referring Image Segmentation.” In [cited by applicant]
Maracani, Andrea, Umberto Michieli, Marco Toldo, and Pietro Zanuttigh. “RECALL: Replay-Based Continual Learning in Semantic Segmentation.” In [cited by applicant]
Michieli, Umberto, and Pietro Zanuttigh. “Continual Semantic Segmentation via Repulsion-Attraction of Sparse and Disentangled Latent Representations.” [cited by applicant]
Michieli, Umberto, and Pietro Zanuttigh. “Incremental Learning Techniques for Semantic Segmentation.” In [cited by applicant]
Michieli, Umberto, and Pietro Zanuttigh. “Knowledge Distillation for Incremental Learning in Semantic Segmentation.” arXiv, Jan. 20, 2021. [cited by applicant]
Mikolov, Tomas, Kai Chen, Greg Corrado, and Jeffrey Dean. “Efficient Estimation of Word Representations in Vector Space.” arXiv, Sep. 6, 2013. [cited by applicant]
Parisi, German I., Ronald Kemker, Jose L. Part, Christopher Kanan, and Stefan Wermter. “Continual Lifelong Learning with Neural Networks: A Review.” arXiv, 2019. [cited by applicant]
Pennington, Jeffrey, Richard Socher, and Christopher Manning. “Glove: Global Vectors for Word Representation.” In [cited by applicant]
Qi, Hang, Matthew Brown, and David G. Lowe. “Low-Shot Learning with Imprinted Weights.” In [cited by applicant]
Qian, Rui, Yunchao Wei, Honghui Shi, Jiachen Li, Jiaying Liu, and Thomas Huang. “Weakly Supervised Scene Parsing with Point-Based Distance Metric Learning.” [cited by applicant]
Roy, Anirban, and Sinisa Todorovic. “Combining Bottom-Up, Top-Down, and Smoothness Cues for Weakly Supervised Image Segmentation.” [cited by applicant]
Sariyildiz, Mert Bulent, Julien Perez, and Diane Larlus. “Learning Visual Representations with Caption Annotations.” In [cited by applicant]
Mert Bulent Sariyildiz, Yannis Kalantidis, Diane Larlus, and Karteek Alahari. Concept generalization in visual representation learning. In ICCV, 2021. [cited by applicant]
Shi, Miaojing, Holger Caesar, and Vittorio Ferrari. “Weakly Supervised Object Localization Using Things and Stuff Transfer.” [cited by applicant]
Siam, Mennatullah, Boris Oreshkin, and Martin Jagersand. “AMP: Adaptive Masked Proxies for Few-Shot Segmentation.” [cited by applicant]
Song, Chunfeng, Yan Huang, Wanli Ouyang, and Liang Wang. “Box-Driven Class-Wise Region Masking and Filling Rate Guided Loss for Weakly Supervised Semantic Segmentation.” [cited by applicant]
Tang, Meng, Abdelaziz Djelouah, Federico Perazzi, Yuri Boykov, and Christopher Schroers. “Normalized Cut Loss for Weakly-Supervised CNN Segmentation.” [cited by applicant]
Tian, Yonglong, Yue Wang, Dilip Krishnan, Joshua B. Tenenbaum, and Phillip Isola. “Rethinking Few-Shot Image Classification: A Good Embedding Is All You Need?” arXiv, Jun. 17, 2020. [cited by applicant]
WeiWang, VincentWZheng, Han Yu, and Chunyan Miao. A survey of zero-shot learning: Settings, methods, and applications. ACM Transactions on Intelligent Systems and Technology (TIST), 2019. [cited by applicant]
Yaqing Wang, Quanming Yao, James T Kwok, and Lionel M Ni. Generalizing from a few examples: A survey on few-shot learning. ACM computing surveys (csur), 2020. [cited by applicant]
Wang, Yude, Jie Zhang, Meina Kan, Shiguang Shan, and Xilin Chen. Self-Supervised Equivariant Attention Mechanism for Weakly Supervised Semantic Segmentation [cited by applicant]
Wang, Zhaoqing, Yu Lu, Qiang Li, Xunqiang Tao, Yandong Guo, Mingming Gong, and Tongliang Liu. “CRIS: CLIP-Driven Referring Image Segmentation.” [cited by applicant]
Wu, Zifeng, Chunhua Shen, and Anton van den Hengel. “Wider or Deeper: Revisiting the ResNet Model for Visual Recognition.” arXiv, Nov. 30, 2016. [cited by applicant]
Xian, Yongqin, Subhabrata Choudhury, Yang He, Bernt Schiele, and Zeynep Akata. “Semantic Projection Network for Zero- and Few-Label Semantic Segmentation.” [cited by applicant]
Xie, Enze, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M Alvarez, and Ping Luo. “SegFormer: Simple and Ef?cient Design for Semantic Segmentation with Transformers,” NeurIPS, 2021. [cited by applicant]
Xu, Jiarui, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, and Xiaolong Wang. “GroupViT: Semantic Segmentation Emerges from Text Supervision.” In [cited by applicant]
Xu, Lian, Wanli Ouyang, Mohammed Bennamoun, Farid Boussaid, and Dan Xu. “Multi-Class Token Transformer for Weakly Supervised Semantic Segmentation.” [cited by applicant]
Zhao, Hang, Xavier Puig, Bolei Zhou, Sanja Fidler, and Antonio Torralba. “Open Vocabulary Scene Parsing.” [cited by applicant]
Zhou, Bolei, Aditya Khosla, Agata Lapedriza, Aude Oliva, and Antonio Torralba. “Learning Deep Features for Discriminative Localization.” [cited by applicant]
Zhou, Bolei, Hang Zhao, Xavier Puig, Sanja Fidler, Adela Barriuso, and Antonio Torralba. “Scene Parsing through ADE20K Dataset.” [cited by applicant]
Brown, Tom, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. Advances in neural… [cited by applicant]
Pietro Buzzega, Matteo Boschini, Angelo Porrello, Davide Abati, and Simone Calderara. Dark experience for general continual learning: a strong, simple baseline. Advances in neural information processing systems, 2020. [cited by applicant]
Liang-Chieh Chen, George Papandreou, Iasonas Kokkinos, Kevin Murphy, and Alan L Yuille. Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs. IEEE transactions … [cited by applicant]
Ali Cheraghian, Shafin Rahman, Pengfei Fang, Soumava Kumar Roy, Lars Petersson, and Mehrtash Harandi. Semantic-aware knowledge distillation for fewshot class-incremental learning. In CVPR, 2021. [cited by applicant]
Csurka, Gabriela, Riccardo Volpi, and Boris Chidlovskii. “Semantic Image Segmentation: Two Decades of Research.” [cited by applicant]
Jia Deng,Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In CVPR, 2009. [cited by applicant]
Everingham, Mark, Luc Van Gool, Christopher K. I. Williams, John Winn, and Andrew Zisserman. “The Pascal Visual Object Classes (VOC) Challenge.” [cited by applicant]
Ghiasi, Golnaz, Xiuye Gu, Yin Cui, and Tsung-Yi Lin. “Scaling Open-Vocabulary Image Segmentation with Image-Level Labels.” In [cited by applicant]
Albert Gordo and Diane Larlus. Beyond instance-level image retrieval: Leveraging captions to learn a global visual representation for semantic retrieval. In CVPR, 2017. [cited by applicant]
Tyler L. Hayes, Nathan D. Cahill, and Christopher Kanan. Memory Efficient Experience Replay for Streaming Learning. In ICRA, 2019. [cited by applicant]
Anna Khoreva, Rodrigo Benenson, Jan Hosang, Matthias Hein, and Bernt Schiele. Simple does it: Weakly supervised instance and semantic segmentation. In CVPR, 2017. [cited by applicant]
Marvin Klingner, Andreas B{umlaut over ( )}ar, Philipp Donn, and Tim Fingscheidt. Class incremental learning for semantic segmentation re-using neither old data nor old labels. In 2020 IEEE 23rd International Conference… [cited by applicant]
Jungbeom Lee, Eunji Kim, Sungmin Lee, Jangho Lee, and Sungroh Yoon. Ficklenet: Weakly and semi-supervised semantic image segmentation using stochastic inference. In CVPR, pp. 5267-5276, 2019. [cited by applicant]
Lee, Seungho, Minhyun Lee, Jongwuk Lee, and Hyunjung Shim. “Railroad Is Not a Train: Saliency as Pseudo-Pixel Supervision for Weakly Supervised Semantic Segmentation.” arXiv, May 19, 2021. [cited by applicant]
David Lopez-Paz and Marc'Aurelio Ranzato. Gradient Episodic Memory for Continual Learning. In NeurIPS, 2017. [cited by applicant]
Masana, Marc, Xialei Liu, Bartlomiej Twardowski, Mikel Menta, Andrew D. Bagdanov, and Joost van de Weijer. “Class-Incremental Learning: Survey and Performance Evaluation on Image Classification.” arXiv, 2021. [cited by applicant]
Michael McCloskey and Neal J Cohen. Catastrophic interference in connectionist networks: The sequential learning problem. In Psychology of learning and motivation. 1989. [cited by applicant]
George A Miller. Wordnet: a lexical database for English. Communications of the ACM, 1995. [cited by applicant]
Quattoni, Ariadna, Michael Collins, and Trevor Darrell. “Learning Visual Representations Using Images with Captions.” In [cited by applicant]
Radford, Alec, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, et al. “Learning Transferable Visual Models From Natural Language Supervision,” ICML, 2021. [cited by applicant]
Sylvestre-Alvise Rebuffi, Alexander Kolesnikov, Georg Sperl, and Christoph H. Lampert. iCaRL: Incremental Classifier and Representation Learning. In CVPR, 2017. [cited by applicant]
Guolei Sun, Wenguan Wang, Jifeng Dai, and Luc Van Gool. Mining cross-image semantics for weakly supervised semantic segmentation. In ECCV, 2020. [cited by applicant]
Yue Wu, Yinpeng Chen, Lijuan Wang, Yuancheng Ye, Zicheng Liu, Yandong Guo, and Yun Fu. Large scale incremental learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 374-382,… [cited by applicant]