ComputerVisionFoundation Videos Cross-Batch Memory for Embedding Learning
🎯 Загружено автоматически через бота:
🚫 Оригинал видео:
📺 Данное видео принадлежит каналу «ComputerVisionFoundation Videos» (@ComputerVisionFoundation). Оно представлено в нашем сообществе исключительно в информационных, научных, образовательных или культурных целях. Наше сообщество не утверждает никаких прав на данное видео. Пожалуйста, поддержите автора, посетив его оригинальный канал.
✉️ Если у вас есть претензии к авторским правам на данное видео, пожалуйста, свяжитесь с нами по почте support@, и мы немедленно удалим его.
📃 Оригинальное описание:
Authors: Xun Wang, Haozhi Zhang, Weilin Huang, Matthew R. Scott Description: Mining informative negative instances are of central importance to deep metric learning (DML). However, the hard-mining ability of existing DML methods is intrinsically limited by mini-batch training, where only a mini-batch of instances are accessible at each iteration. In this paper, we identify a “slow drift” phenomena by observing that the embedding features drift exceptionally slow even as the model parameters are updating throughout the training process. It suggests that the features of instances computed at preceding iterations can considerably approximate to their features extracted by current model. We propose a cross-batch memory (XBM) mechanism that memorizes the embeddings of past iterations, allowing the model to collect sufficient hard negative pairs across multiple mini-batches - even over the whole dataset. Our XBM can be directly integrated into general pair-based DML demonstrate that, without bells and whistles, XBM augmented DML can boost the performance considerably on image retrieval. In particular, with XBM, a simple contrastive loss can have large R@1 improvements of 12%-22.5% on three large-scale datasets, easily surpassing the most sophisticated state-of-the-art methods [38, 27, 2], by a large margin. Our XBM is conceptually simple, easy to implement - using several lines of codes, and is memory efficient - with a negligible 0.2 GB extra GPU memory.
5 views
10
4
1 day ago 00:04:48 10
[ComputerVisionFoundation Videos] Cross-Batch Memory for Embedding Learning
1 month ago 00:30:55 1
Генерация хайлайтов / Евгений Россинский (ivi)
3 months ago 00:11:30 1
DARPA - robots and technologies for the future management of advanced US research | PRO Robots
5 months ago 00:17:15 1
AxOS Plasma Edition
9 months ago 00:05:15 1
Betty Holberton: A Pioneer of Computer Programming
11 months ago 01:17:20 1
Most Self-Help Advice Is Wrong. Here’s The Fastest Way To Transform Your Life | Cal Newport
11 months ago 00:02:31 1
Download & Install Blender 4.0 | How to download and Install Blender | CADable | CADable tutorials
1 year ago 00:10:26 1
Autodistill: Train YOLOv8 with ZERO Annotations
1 year ago 00:32:03 6
The Art of Procedural Noise #SoME3
1 year ago 01:28:30 4
Assembly, System Calls, and Hardware in C++ - David Sankel - CppNow 2023
1 year ago 00:15:08 3
A New Arch Linux Update is Available
2 years ago 00:14:49 1
Should we fear AI?
2 years ago 00:02:47 1
Augmented Reality with X-Ray Vision
2 years ago 01:38:15 3
AI for nuclear energy | AI FOR GOOD WEBINARS
2 years ago 00:06:29 1
What is deep learning and why is it a data-driven approach?
3 years ago 01:22:28 1
3Lux - techno video mix (1991)
3 years ago 01:10:26 1
3Lux-2 - the non stop techno trance video mix (1992)
5 years ago 00:14:30 22
OpenCV Object Detection in Games Python Tutorial #1
8 years ago 00:02:18 17
Rainforest Spiritual Enslavement - Homes Built Over The Sea (video by Soul Archive)