AI Paper Atlas

Research map / Video / Action recognition

What Would You Expect? Anticipating Egocentric Actions With Rolling-Unrolling LSTMs and Modality Attention

Antonino Furnari, Giovanni Maria Farinella

ICCV 2019 · Video › Action recognition · approach: Disentangled action recognition · 183 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
replace single-LSTM late-fusion baseline by rolling-unrolling LSTMs (late fusion) (ablation)BL · EPIC-Kitchens val (anticipation, tau_a=1s), ablationTop-5 action accuracy %: 32.02 -> 34.23Improved
replace late fusion by Modality ATTention (MATT) (ablation)RU · EPIC-Kitchens val (anticipation, tau_a=1s), ablationTop-5 action accuracy %: 34.23 -> 35.32Improved
replace early fusion by MATT (ablation)RU · EPIC-Kitchens val (anticipation, tau_a=1s), ablationTop-5 action accuracy %: 31.88 -> 35.32Improved
fuse three modalities instead of RGB only (ablation)RU · EPIC-Kitchens val (anticipation, tau_a=1s), ablationTop-5 action accuracy %: 30.83 -> 35.32Improved
add sequence completion pre-training (ablation)RU · EPIC-Kitchens val (anticipation, tau_a=1s), ablationTop-5 action accuracy %: 34.75 -> 35.32Improved
add sequence completion pre-training at tau_a=1.5s (ablation)RU · EPIC-Kitchens val (anticipation, tau_a=1s), ablationTop-5 action accuracy %: 32.34 -> 32.24No clear effect
full RU with MATT and pre-training vs strong single-LSTM baseline (ablation)RU · EPIC-Kitchens val (anticipation, tau_a=1s), ablationTop-5 action accuracy %: 32.02 -> 35.32Improved
use RU instead of best prior method on EPIC valRU vs RL · EPIC-Kitchens val (tau_a=1s)Top-5 action accuracy %: 29.61 -> 35.32Improved
use RU on EGTEA Gaze+ instead of ELRU vs EL · EGTEA Gaze+ (tau_a=1s)Top-5 action accuracy %: 64.62 -> 66.40Improved
use RU for action recognition on EGTEA vs LSTARU vs LSTA · EGTEA Gaze+ (action recognition)accuracy %: 61.86 -> 60.20Hurt
RU for verb top-1 recognition on EPIC test S1 vs LSTARU vs LSTA · EPIC-Kitchens test S1 (recognition)Top-1 verb accuracy %: 59.55 -> 56.93Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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