Foreground-Background Separation through Concept Distillation from Generative Image Foundation Models

被引:3
|
作者
Dombrowski, Mischa [1 ]
Reynaud, Hadrien [2 ]
Baugh, Matthew [2 ]
Kainz, Bernhard [1 ,2 ]
机构
[1] Friedrich Alexander Univ Erlangen Nurnberg, Nurnberg, Germany
[2] Imperial Coll London, London, England
来源
2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION, ICCV | 2023年
关键词
D O I
10.1109/ICCV51070.2023.00097
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Curating datasets for object segmentation is a difficult task. With the advent of large-scale pre-trained generative models, conditional image generation has been given a significant boost in result quality and ease of use. In this paper, we present a novel method that enables the generation of general foreground-background segmentation models from simple textual descriptions, without requiring segmentation labels. We leverage and explore pre-trained latent diffusion models, to automatically generate weak segmentation masks for concepts and objects. The masks are then used to fine-tune the diffusion model on an inpainting task, which enables fine-grained removal of the object, while at the same time providing a synthetic foreground and background dataset. We demonstrate that using this method beats previous methods in both discriminative and generative performance and closes the gap with fully supervised training while requiring no pixel-wise object labels. We show results on the task of segmenting four different objects (humans, dogs, cars, birds) and a use case scenario in medical image analysis. The code is available at https://github.com/MischaD/fobadiffusion.
引用
收藏
页码:988 / 998
页数:11
相关论文
共 50 条
  • [31] Foreground-Background Contrastive Learning for Few-Shot Remote Sensing Image Scene Classification
    Geng, Jie
    Xue, Bohan
    Jiang, Wen
    IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2023, 61
  • [32] Integrating foreground-background feature distillation and contrastive feature learning for ultra-fine-grained visual classification
    Chen, Qiupu
    Jiao, Lin
    Wang, Fenmei
    Du, Jianming
    Liu, Haiyun
    Wang, Xue
    Wang, Rujing
    PATTERN RECOGNITION, 2024, 150
  • [33] Background and foreground disentangled generative adversarial network for scene image synthesis
    Ni, Jiancheng
    Zhang, Susu
    Zhou, Zili
    Hou, Lijun
    Hou, Jie
    Gao, Feng
    COMPUTERS & GRAPHICS-UK, 2021, 97 : 54 - 66
  • [34] Background modeling for generative image models
    Schoenborn, Sandro
    Egger, Bernhard
    Forster, Andreas
    Vetter, Thomas
    COMPUTER VISION AND IMAGE UNDERSTANDING, 2015, 136 : 117 - 127
  • [35] Foreground-Background Partitioning and Feature Fusion for Weakly Supervised Fine-Grained Image Recognition
    Liu, Chenglin
    Li, Jiuliang
    Chen, Yanmin
    Luo, Jun
    Zhou, Mengyao
    Yang, Jian
    Li, Zhenfei
    PATTERN RECOGNITION AND COMPUTER VISION, PT III, PRCV 2024, 2025, 15033 : 17 - 30
  • [36] A New Foreground-Background based Method for Behavior-Oriented Social Media Image Classification
    Nandanwar, Lokesh
    Shivakumara, Palaiahnakote
    Krishnani, Divya
    Ramachandra, Raghavendra
    Lu, Tong
    Pal, Umapada
    Kankanhalli, Mohan
    ACM TRANSACTIONS ON MULTIMEDIA COMPUTING COMMUNICATIONS AND APPLICATIONS, 2021, 17 (04)
  • [37] Occlusion Handling by Successively Excluding Foregrounds for Light Field Depth Estimation Based on Foreground-Background Separation
    Lee, Jae Young
    Park, Rae-Hong
    Kim, Junmo
    IEEE ACCESS, 2021, 9 : 103927 - 103936
  • [38] Multi-scale foreground-background separation for light field depth estimation with deep convolutional networks
    Lee, Jae Young
    Hur, Jiwan
    Choi, Jaehyun
    Park, Rae-Hong
    Kim, Junmo
    PATTERN RECOGNITION LETTERS, 2023, 171 : 138 - 147
  • [39] Video Foreground-Background Separation via Weighted Schatten-p Norm and Structured Sparsity Decomposition
    Wei Yufeng
    Jing Mingli
    Li Lan
    Sun Kun
    Fan Ruibo
    LASER & OPTOELECTRONICS PROGRESS, 2021, 58 (08)
  • [40] LRSDSFD: low-rank sparse decomposition and symmetrical frame difference method for moving video foreground-background separation
    Gao, Hongqiao
    EAI ENDORSED TRANSACTIONS ON SCALABLE INFORMATION SYSTEMS, 2022, 9 (36)