AI Paper Atlas

Research map / Video / Video segmentation and tracking

Labelling unlabelled videos from scratch with multi-modal self-supervision

Yuki Asano, Mandela Patrick, Christian Rupprecht, Andrea Vedaldi

NeurIPS 2020 · Video › Video segmentation and tracking · approach: Self-supervised video representation learning · 80 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
Train clusters and representations jointly with SeLaVi instead of applying k-means to pretrained representationsVGG-Sound video clustering · VGG-SoundNMI: 48.5 -> 55.9Improved
Train clusters and representations jointly with SeLaVi instead of applying k-means to pretrained representationsAVE video clustering · AVENMI: 56.3 -> 66.2Improved
Train clusters and representations jointly with SeLaVi instead of applying k-means to pretrained representationsKinetics video clustering · KineticsNMI: 48.9 -> 27.1Hurt
Train clusters and representations jointly with SeLaVi instead of applying k-means to pretrained representationsKinetics-Sound video clustering · Kinetics-SoundNMI: 47.5 -> 47.5No clear effect
Add audio to visual input (ablation)SeLaVi · VGG-SoundNMI: 52.8 -> 55.9Improved
Add visual input to audio-only clustering (ablation)SeLaVi · VGG-SoundNMI: 47.5 -> 55.9Improved
Use modality-splicing and shared multi-modal labels instead of feature concatenation (ablation)SeLaVi · VGG-SoundNMI: 24.7 -> 48.8Improved
Add modality alignment (ablation)SeLaVi · VGG-SoundNMI: 50.9 -> 51.1Improved
Add Gaussian marginals (ablation)SeLaVi · VGG-SoundNMI: 51.5 -> 51.1Hurt
Add decorrelated heads (ablation)SeLaVi · VGG-SoundNMI: 49.9 -> 51.1Improved
Vary the number of clusters from 309 to 256 or 619 (ablation)SeLaVi · VGG-SoundNMI: 56.7 -> 56.8; 56.7 -> 56.9No clear effect
Increase the number of clusters from 309 to 619 (ablation)SeLaVi · VGG-Soundmean maximal purity: 38.0 -> 42.7Improved

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