Low-Cost Training of Image-to-Image Diffusion Models with Incremental Learning and Task/Domain Adaptation

被引:0
|
作者
Antona, Hector [1 ]
Otero, Beatriz [1 ]
Tous, Ruben [1 ]
机构
[1] Univ Politecn Cataluna, Dept Comp Architecture, Jordi Girona 1-3, Barcelona 08034, Spain
关键词
diffusion probabilistic models; deep learning; adaptive learning; transfer learning; image inpainting; image colorization; image-to-image translation; training efficiency;
D O I
10.3390/electronics13040722
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Diffusion models specialized in image-to-image translation tasks, like inpainting and colorization, have outperformed the state of the art, yet their computational requirements are exceptionally demanding. This study analyzes different strategies to train image-to-image diffusion models in a low-resource setting. The studied strategies include incremental learning and task/domain transfer learning. First, a base model for human face inpainting is trained from scratch with an incremental learning strategy. The resulting model achieves an FID score almost equivalent to that of its batch learning equivalent while significantly reducing the training time. Second, the base model is fine-tuned to perform a different task, image colorization, and, in a different domain, landscape images. The resulting colorization models showcase exceptional performances with a minimal number of training epochs. We examine the impact of different configurations and provide insights into the ability of image-to-image diffusion models for transfer learning across tasks and domains.
引用
收藏
页数:21
相关论文
共 50 条
  • [31] Low-cost image indexing for massive database
    Je-Ho Park
    Multimedia Tools and Applications, 2015, 74 : 2237 - 2255
  • [32] LOW-COST SONAR IMAGE-PROCESSING
    JOURDAN, DW
    SEA TECHNOLOGY, 1995, 36 (08) : 45 - 47
  • [33] A Low-Cost Pavement Image Acquisition System
    Chiculita, Claudiu
    Frangu, Laurentiu
    2016 IEEE 22ND INTERNATIONAL SYMPOSIUM FOR DESIGN AND TECHNOLOGY AND ELECTRONIC PACKAGING (SIITME), 2016, : 279 - 282
  • [34] LOW-COST INTERACTIVE IMAGE PROCESSING.
    Johnston, Emily G.
    Rosenfeld, Azriel
    Proceedings of the Society for Information Display, 1975, 16 (01): : 1 - 7
  • [35] Unsupervised Domain Adaptation for Medical Image Segmentation by Disentanglement Learning and Self-Training
    Xie, Qingsong
    Li, Yuexiang
    He, Nanjun
    Ning, Munan
    Ma, Kai
    Wang, Guoxing
    Lian, Yong
    Zheng, Yefeng
    IEEE TRANSACTIONS ON MEDICAL IMAGING, 2024, 43 (01) : 4 - 14
  • [36] A Low-Cost Pedagogical Environment for Training on Technologies for Image-Guided Robotic Surgery
    Rodriguez-Vila, B.
    Gutierrez, A.
    Peral-Boiza, M.
    Ying, H.
    Gomez-Fernandez, T.
    Gomez, E. J.
    Sanchez-Gonzalez, P.
    WORLD CONGRESS ON MEDICAL PHYSICS AND BIOMEDICAL ENGINEERING 2018, VOL 2, 2019, 68 (02): : 821 - 824
  • [37] A Low-Cost, Passive Navigation Training System for Image-Guided Spinal Intervention
    Lorias-Espinoza, Daniel
    Gonzalez Carranza, Vicente
    Chico-Ponce de Leon, Fernando
    Perez Escamirosa, Fernando
    Minor Martinez, Arturo
    WORLD NEUROSURGERY, 2016, 95 : 322 - 328
  • [38] Unsupervised Domain Adaptation for the Semantic Segmentation of Remote Sensing Images via One-Shot Image-to-Image Translation
    Ismael, Sarmad F.
    Kayabol, Koray
    Aptoula, Erchan
    IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, 2023, 20
  • [39] CROSS DOMAIN ADAPTATION FOR ON-ROAD OBJECT DETECTION USING MULTIMODAL STRUCTURE-CONSISTENT IMAGE-TO-IMAGE TRANSLATION
    Lin, Che-Tsung
    2019 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), 2019, : 3029 - 3030
  • [40] Design and Implementation of a Low-cost Guided Image Filter for Underwater Image Enhancement
    Chang, Cheng-Hao
    Huang, I-Chang
    Lin, Mu-Fan
    Kuang, Shiann-Rong
    2017 IEEE CONFERENCE ON DEPENDABLE AND SECURE COMPUTING, 2017, : 296 - 299