BloomCoreset: Fast Coreset Sampling using Bloom Filters for Fine-Grained Self-Supervised Learning

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Singh, Prajwal, Vashishtha, Gautam, Mastan, Indra Deep, Raman, Shanmuganathan
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913622910828544
author Singh, Prajwal
Vashishtha, Gautam
Mastan, Indra Deep
Raman, Shanmuganathan
author_facet Singh, Prajwal
Vashishtha, Gautam
Mastan, Indra Deep
Raman, Shanmuganathan
contents The success of deep learning in supervised fine-grained recognition for domain-specific tasks relies heavily on expert annotations. The Open-Set for fine-grained Self-Supervised Learning (SSL) problem aims to enhance performance on downstream tasks by strategically sampling a subset of images (the Core-Set) from a large pool of unlabeled data (the Open-Set). In this paper, we propose a novel method, BloomCoreset, that significantly reduces sampling time from Open-Set while preserving the quality of samples in the coreset. To achieve this, we utilize Bloom filters as an innovative hashing mechanism to store both low- and high-level features of the fine-grained dataset, as captured by Open-CLIP, in a space-efficient manner that enables rapid retrieval of the coreset from the Open-Set. To show the effectiveness of the sampled coreset, we integrate the proposed method into the state-of-the-art fine-grained SSL framework, SimCore [1]. The proposed algorithm drastically outperforms the sampling strategy of the baseline in SimCore [1] with a $98.5\%$ reduction in sampling time with a mere $0.83\%$ average trade-off in accuracy calculated across $11$ downstream datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2412_16942
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle BloomCoreset: Fast Coreset Sampling using Bloom Filters for Fine-Grained Self-Supervised Learning
Singh, Prajwal
Vashishtha, Gautam
Mastan, Indra Deep
Raman, Shanmuganathan
Computer Vision and Pattern Recognition
The success of deep learning in supervised fine-grained recognition for domain-specific tasks relies heavily on expert annotations. The Open-Set for fine-grained Self-Supervised Learning (SSL) problem aims to enhance performance on downstream tasks by strategically sampling a subset of images (the Core-Set) from a large pool of unlabeled data (the Open-Set). In this paper, we propose a novel method, BloomCoreset, that significantly reduces sampling time from Open-Set while preserving the quality of samples in the coreset. To achieve this, we utilize Bloom filters as an innovative hashing mechanism to store both low- and high-level features of the fine-grained dataset, as captured by Open-CLIP, in a space-efficient manner that enables rapid retrieval of the coreset from the Open-Set. To show the effectiveness of the sampled coreset, we integrate the proposed method into the state-of-the-art fine-grained SSL framework, SimCore [1]. The proposed algorithm drastically outperforms the sampling strategy of the baseline in SimCore [1] with a $98.5\%$ reduction in sampling time with a mere $0.83\%$ average trade-off in accuracy calculated across $11$ downstream datasets.
title BloomCoreset: Fast Coreset Sampling using Bloom Filters for Fine-Grained Self-Supervised Learning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2412.16942