ProtoDepth: Unsupervised Continual Depth Completion with Prototypes

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Rim, Patrick, Park, Hyoungseob, Gangopadhyay, S., Zeng, Ziyao, Chung, Younjoon, Wong, Alex
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912277467234304
author Rim, Patrick
Park, Hyoungseob
Gangopadhyay, S.
Zeng, Ziyao
Chung, Younjoon
Wong, Alex
author_facet Rim, Patrick
Park, Hyoungseob
Gangopadhyay, S.
Zeng, Ziyao
Chung, Younjoon
Wong, Alex
contents We present ProtoDepth, a novel prototype-based approach for continual learning of unsupervised depth completion, the multimodal 3D reconstruction task of predicting dense depth maps from RGB images and sparse point clouds. The unsupervised learning paradigm is well-suited for continual learning, as ground truth is not needed. However, when training on new non-stationary distributions, depth completion models will catastrophically forget previously learned information. We address forgetting by learning prototype sets that adapt the latent features of a frozen pretrained model to new domains. Since the original weights are not modified, ProtoDepth does not forget when test-time domain identity is known. To extend ProtoDepth to the challenging setting where the test-time domain identity is withheld, we propose to learn domain descriptors that enable the model to select the appropriate prototype set for inference. We evaluate ProtoDepth on benchmark dataset sequences, where we reduce forgetting compared to baselines by 52.2% for indoor and 53.2% for outdoor to achieve the state of the art.
format Preprint
id arxiv_https___arxiv_org_abs_2503_12745
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ProtoDepth: Unsupervised Continual Depth Completion with Prototypes
Rim, Patrick
Park, Hyoungseob
Gangopadhyay, S.
Zeng, Ziyao
Chung, Younjoon
Wong, Alex
Computer Vision and Pattern Recognition
We present ProtoDepth, a novel prototype-based approach for continual learning of unsupervised depth completion, the multimodal 3D reconstruction task of predicting dense depth maps from RGB images and sparse point clouds. The unsupervised learning paradigm is well-suited for continual learning, as ground truth is not needed. However, when training on new non-stationary distributions, depth completion models will catastrophically forget previously learned information. We address forgetting by learning prototype sets that adapt the latent features of a frozen pretrained model to new domains. Since the original weights are not modified, ProtoDepth does not forget when test-time domain identity is known. To extend ProtoDepth to the challenging setting where the test-time domain identity is withheld, we propose to learn domain descriptors that enable the model to select the appropriate prototype set for inference. We evaluate ProtoDepth on benchmark dataset sequences, where we reduce forgetting compared to baselines by 52.2% for indoor and 53.2% for outdoor to achieve the state of the art.
title ProtoDepth: Unsupervised Continual Depth Completion with Prototypes
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2503.12745