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
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
  • [1] Unsupervised Ensemble Semantic Segmentation for Foreground-Background Separation on Satellite Image
    Tarry, Jaelen
    Dong, Xishuang
    Li, Xiangfang
    Qian, Lijun
    Chance, Leah
    Morrone, Philip
    18TH IEEE INTERNATIONAL CONFERENCE ON SEMANTIC COMPUTING, ICSC 2024, 2024, : 212 - 217
  • [2] Foreground-Background Ambient Sound Scene Separation
    Olvera, Michel
    Vincent, Emmanuel
    Serizel, Romain
    Gasso, Gilles
    28TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO 2020), 2021, : 281 - 285
  • [3] Foreground-background separation technique for crack detection
    Nayyeri, Fereshteh
    Hou, Lei
    Zhou, Jun
    Guan, Hong
    COMPUTER-AIDED CIVIL AND INFRASTRUCTURE ENGINEERING, 2019, 34 (06) : 457 - 470
  • [4] Image Foreground-Background Separation Based on Texture Features Extracted in Lab Color Space
    Yang Chao
    Liu Benyong
    LASER & OPTOELECTRONICS PROGRESS, 2019, 56 (12)
  • [5] Aesthetic-aware image retargeting based on foreground-background separation and PSO optimization
    Naderi, Mohammad Reza
    Givkashi, Mohammad Hossein
    Karimi, Nader
    Shirani, Shahram
    Samavi, Shadrokh
    MULTIMEDIA TOOLS AND APPLICATIONS, 2023, 83 (12) : 34867 - 34886
  • [6] On foreground-background separation in low quality color document images
    Garain, U
    Paquet, T
    Heutte, L
    EIGHTH INTERNATIONAL CONFERENCE ON DOCUMENT ANALYSIS AND RECOGNITION, VOLS 1 AND 2, PROCEEDINGS, 2005, : 585 - 589
  • [7] Foreground-background separation and deblurring super-resolution method☆
    Liu, Xuebin
    Chen, Yuang
    Zhao, Chongji
    Yang, Jie
    Deng, Huan
    OPTICS AND LASERS IN ENGINEERING, 2025, 184
  • [8] An Adaptive Foreground-Background Separation Method for Effective Binarization of Document Images
    Das, Bishwadeep
    Bhowmik, Showmik
    Saha, Aniruddha
    Sarkar, Ram
    PROCEEDINGS OF THE EIGHTH INTERNATIONAL CONFERENCE ON SOFT COMPUTING AND PATTERN RECOGNITION (SOCPAR 2016), 2018, 614 : 515 - 524
  • [9] Unified graph-based method for instance separation from foreground-background segmentation
    Spasc, Milica
    Mihajlovc, Igor
    Spasc, Nikola
    Jankovc, Dragan
    2022 57TH INTERNATIONAL SCIENTIFIC CONFERENCE ON INFORMATION, COMMUNICATION AND ENERGY SYSTEMS AND TECHNOLOGIES (ICEST), 2022, : 115 - 118
  • [10] An algorithm for foreground-background separation in low quality patrimonial document images
    Mello, Carlos A. B.
    PROGRESS IN PATTERN RECOGNITION, IMAGE ANALYSIS AND APPLICATIONS, PROCEEDINGS, 2007, 4756 : 911 - 920