</> Code with Papers by Syntology

Papers › Training for Diversity in Image Paragraph Captioning

Training for Diversity in Image Paragraph Captioning

1 Oct 2018EMNLP 2018 10archive 2025-07-28

Luke Melas-Kyriazi, Alex Rush, er, George Han

Image paragraph captioning models aim to produce detailed descriptions of a source image. These models use similar techniques as standard image captioning models, but they have encountered issues in text generation, notably a lack of diversity between sentences, that have limited their effectiveness. In this work, we consider applying sequence-level training for this task. We find that standard self-critical training produces poor results, but when combined with an integrated penalty on trigram repetition produces much more diverse paragraphs. This simple training approach improves on the best result on the Visual Genome paragraph captioning dataset from 16.9 to 30.6 CIDEr, with gains on METEOR and BLEU as well, without requiring any architectural changes.

PaperPDFCode

Code

lukemelas/image-paragraph-captioning mentioned in paperpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

DiversityImage CaptioningImage Paragraph CaptioningMachine TranslationObject DetectionPolicy Gradient MethodsText Generation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Paragraph Captioning Image Paragraph Captioning SCST training, w/ rep. penalty BLEU-4 10.58 #2 of 10 Archive leaderboard report
Image Paragraph Captioning Image Paragraph Captioning SCST training, w/ rep. penalty CIDEr 30.63 #2 of 10 Archive leaderboard report
Image Paragraph Captioning Image Paragraph Captioning SCST training, w/ rep. penalty METEOR 17.86 #2 of 10 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections