AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

Preserving Semantic Neighborhoods for Robust Cross-modal Retrieval

Christopher Thomas, Adriana Kovashka

ECCV 2020 · Vision-language and multimodal › Cross-modal learning · approach: Probabilistic embeddings for cross-modal retrieval · 34 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 within-modality text and image neighbor losses (L_text, L_img)Ang+NP+Sym · PoliticsRecall@1 image-to-text: 0.6270 -> 0.6467Improved
add within-modality text and image neighbor lossesAng+NP+Sym · GoodNewsRecall@1 image-to-text: 0.8704 -> 0.8849Improved
add within-modality text and image neighbor lossesAng+NP+Sym · Conceptual CaptionsRecall@1 image-to-text: 0.7687 -> 0.7760Improved
add within-modality text and image neighbor lossesAng+NP+Sym · COCORecall@1 image-to-text: 0.6976 -> 0.6900Hurt
add the proposed losses to PVSE backbonePVSE · GoodNewsRecall@1 image-to-text: 0.8724 -> 0.8867Improved
add the proposed losses to PVSE backbonePVSE · Flickr30KRecall@1 image-to-text: 0.5539 -> 0.5559No clear effect
add the proposed losses to triplet-loss baselineTrip+NP+Sym · PoliticsRecall@1 image-to-text: 0.4742 -> 0.4940Improved
add the proposed losses to triplet-loss baselineTrip+NP+Sym · COCORecall@1 image-to-text: 0.4957 -> 0.4790Hurt
compute neighbors in Doc2Vec text space instead of image feature space (ablation)Ours · PoliticsRecall@1 image-to-text: 0.6370 -> 0.6467Improved
add L_text (ablation)Ours (Ang) · GoodNewsRecall@1 image-to-text: 0.8786 -> 0.8849Improved
add L_img (ablation)Ours (Ang) · PoliticsRecall@1 image-to-text: 0.6390 -> 0.6467Improved
add the proposed losses to preserve Doc2Vec neighborhoods in joint spaceAng+NP+Sym · Politics (text neighborhood preservation)neighbor overlap (text): 0.1544 -> 0.1703Improved

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