AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

AESOP: Abstract Encoding of Stories, Objects, and Pictures

Hareesh Ravi, Kushal Kafle, Scott Cohen, Jonathan Brandt, Mubbasir Kapadia

ICCV 2021 · Vision-language and multimodal › Human and embodied modeling · approach: Human perception modeling · 13 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
add text modality (compare proposed multimodal model to unimodal visual model) (ablation)AESOP model · AESOP Assistant IllustratorScene similarity: 4.2 -> 4.1No clear effect
add story context (compare to one-to-one model without story context) (ablation)AESOP model · AESOP Assistant IllustratorScene similarity: 1.2 -> 4.1Improved
use abstract object representation instead of pretrained ResNet-18 pixel model (ablation)AESOP model · AESOP Assistant IllustratorScene similarity: 1.0 -> 4.1Improved
add visual input to writer (multimodal vs GPT-2 unimodal) (ablation)AESOP model · AESOP Assistant WriterBLEU-4: 0.41 -> 1.96Improved
add visual input to writer (multimodal vs GPT-2 unimodal) (ablation)AESOP model · AESOP Assistant WriterCIDEr: 5.7 -> 17.9Improved
add story context to writer (vs one-to-one) (ablation)AESOP model · AESOP Assistant WriterCIDEr: 7.2 -> 17.9Improved
compare proposed model against human baseline on writer metricsAESOP model · AESOP Assistant WriterMETEOR: 11.1 -> 9.02Hurt
compare proposed model against human baseline on writer metricsAESOP model · AESOP Assistant WriterBLEU-4: 1.84 -> 1.96Improved
copy previous panel (Repeat) versus proposed modelRepeat baseline · AESOP Assistant IllustratorScene similarity: 5.1 -> 4.1Hurt
use AESOP creation process instead of VISTVIST · Panel-swap human testidentification accuracy: 65.8% -> 95%Improved

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