AI Paper Atlas

Research map / Video / Action recognition

Vi2CLR: Video and Image for Visual Contrastive Learning of Representation

Ali Diba, Vivek Sharma, Reza Safdari, Dariush Lotfi, Saquib Sarfraz, Rainer Stiefelhagen, Luc Van Gool

ICCV 2021 · Video › Action recognition · approach: Contrastive video representation learning · 55 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
use FINCH clustering instead of K-means (ablation)Vi2CLR · ImageNet linear eval, 200 epochstop-1 accuracy (%): 73.7 -> 74.3Improved
use FINCH instead of K-means (ablation)Vi2CLR · ImageNet linear eval, 50 epochstop-1 accuracy (%): 64.3 -> 65.5Improved
add video and CenterNCE losses to image-only loss (full Vi2CLR) (ablation)Vi2CLR · ImageNet linear evaltop-1 accuracy (%): 71.2 -> 74.6Improved
add image and video losses to CenterNCE only (ablation)Vi2CLR · ImageNet linear evaltop-1 accuracy (%): 72.7 -> 74.6Improved
add CenterNCE and image loss to video-only loss (ablation)Vi2CLR · UCF101accuracy (%): 86.6 -> 89.1Improved
add full objective vs CenterNCE only (ablation)Vi2CLR · UCF101accuracy (%): 87.8 -> 89.1Improved
add full objective vs video loss only (ablation)Vi2CLR · HMDB51accuracy (%): 51.2 -> 55.7Improved
add full objective vs CenterNCE only (ablation)Vi2CLR · VOC07accuracy (%): 88 -> 89.4Improved
use Vi2CLR instead of SwAVVi2CLR · ImageNet linear evaltop-1 accuracy (%): 74.2 -> 74.6Improved
use Vi2CLR with 300 epochs instead of SwAV 800 epochsVi2CLR · ImageNet 10% labelstop-1 accuracy (%): 70.2 -> 69.7Hurt
use Vi2CLR instead of PCLVi2CLR · VOC07 few-shot k=16accuracy (%): 78.3 -> 80.6Improved
use Vi2CLR instead of PCLVi2CLR · ImageNet 10% labelstop-1 accuracy (%): 75.3 -> 69.7Hurt

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