AI Paper Atlas

Research map / Video / Video segmentation and tracking

Federated Self-Supervised Learning for Video Understanding

Yasar Abbas Ur Rehman, Yan Gao, Jiajun Shen, Pedro Porto Buarque de Gusmão, Nicholas Lane

ECCV 2022 · Video › Video segmentation and tracking · approach: Self-supervised video representation learning · 17 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
pretrain Speed with federated learning instead of centralizedSpeed · UCF-101 clip retrievalR@1: 16.84 -> 21.97Improved
pretrain CtP with federated learning instead of centralizedCtP · UCF-101 action recognition fine-tunetop-1 accuracy: 86.20 -> 81.95Hurt
pretrain CtP with federated learning instead of centralizedCtP · HMDB-51 clip retrievalR@1: 11.80 -> 13.66Improved
use Non-IID (8 classes per client) partition instead of IIDCtP federated · UCF-101 fine-tunetop-1 accuracy: 81.92 -> 81.95No clear effect
use Non-IID (4 classes per client) partition instead of IIDCtP federated · UCF-101 fine-tunetop-1 accuracy: 81.92 -> 81.15No clear effect
use FedU instead of FedAvg aggregationCtP federated · UCF-101 clip retrievalR@1: 32.62 -> 34.07Improved
use loss-based instead of FedAvg aggregationCtP federated · HMDB-51 clip retrievalR@1: 16.54 -> 14.44Hurt
aggregate only backbone weights (FedVSSL alpha=0, beta=0) instead of full model (ablation)FedAvg · UCF-101 clip retrievalR@1: 32.62 -> 34.34Improved
add stochastic weight averaging (beta=1) to loss-weighted aggregation alpha=0 (ablation)FedVSSL · UCF-101 clip retrievalR@1: 34.34 -> 35.61Improved
mix loss and FedAvg aggregation with alpha=0.9 (beta=0) instead of alpha=0 (ablation)FedVSSL · UCF-101 linear probetop-1 accuracy: 47.95 -> 50.36Improved
add alpha=0.9, beta=0 FedVSSL over FedAvg baseline on UCF fine-tuningFedAvg · UCF-101 fine-tunetop-1 accuracy: 79.91 -> 80.62Improved
add SWA (beta=1) to alpha=0.9 FedVSSL (ablation)FedVSSL · UCF-101 fine-tunetop-1 accuracy: 80.62 -> 79.41Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy