Papers › L-Verse: Bidirectional Generation Between Image and Text

L-Verse: Bidirectional Generation Between Image and Text

22 Nov 2021CVPR 2022 1arXiv:2111.11133archive 2025-07-28

TaeHoon Kim, Gwangmo Song, Sihaeng Lee, Sangyun Kim, Yewon Seo, Soonyoung Lee, Seung Hwan Kim, Honglak Lee, Kyunghoon Bae

Far beyond learning long-range interactions of natural language, transformers are becoming the de-facto standard for many vision tasks with their power and scalability. Especially with cross-modal tasks between image and text, vector quantized variational autoencoders (VQ-VAEs) are widely used to make a raw RGB image into a sequence of feature vectors. To better leverage the correlation between image and text, we propose L-Verse, a novel architecture consisting of feature-augmented variational autoencoder (AugVAE) and bidirectional auto-regressive transformer (BiART) for image-to-text and text-to-image generation. Our AugVAE shows the state-of-the-art reconstruction performance on ImageNet1K validation set, along with the robustness to unseen images in the wild. Unlike other models, BiART can distinguish between image (or text) as a conditional reference and a generation target. L-Verse can be directly used for image-to-text or text-to-image generation without any finetuning or extra object detection framework. In quantitative and qualitative experiments, L-Verse shows impressive results against previous methods in both image-to-text and text-to-image generation on MS-COCO Captions. We furthermore assess the scalability of L-Verse architecture on Conceptual Captions and present the initial result of bidirectional vision-language representation learning on general domain.

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

tgisaturday/L-Verse officialmentioned in papermentioned on GitHubpytorch 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

Image CaptioningImage GenerationImage ReconstructionImage to textObject DetectionRepresentation LearningText GenerationText to Image GenerationText-to-Image GenerationZero-Shot Text-to-Image Generationobject-detection

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Captioning COCO Captions L-Verse BLEU-4 39.9 #20 of 41 Archive leaderboard report
Image Captioning COCO Captions L-Verse METEOR 31.4 #20 of 41 Archive leaderboard report
Image Captioning COCO Captions L-Verse ROUGE-L 60.4 #20 of 41 Archive leaderboard report
Image Captioning COCO Captions L-Verse SPICE 23.3 #20 of 41 Archive leaderboard report
Image Reconstruction ImageNet 256x256 AugVAE-ML FID 1.04 #1 of 2 Archive leaderboard report
Image Reconstruction ImageNet 256x256 AugVAE-SL FID 3.28 #2 of 2 Archive leaderboard report
Text-to-Image Generation COCO (Common Objects in Context) L-Verse-CC FID 37.2 #63 of 69 Archive leaderboard report
Text-to-Image Generation COCO (Common Objects in Context) L-Verse-CC FID-1 31.6 #63 of 69 Archive leaderboard report
Text-to-Image Generation COCO (Common Objects in Context) L-Verse-CC FID-2 25.7 #63 of 69 Archive leaderboard report
Text-to-Image Generation COCO (Common Objects in Context) L-Verse-CC FID-4 21.4 #63 of 69 Archive leaderboard report
Text-to-Image Generation COCO (Common Objects in Context) L-Verse-CC FID-8 21.1 #63 of 69 Archive leaderboard report
Text-to-Image Generation COCO (Common Objects in Context) L-Verse FID 45.8 #64 of 69 Archive leaderboard report
Text-to-Image Generation COCO (Common Objects in Context) L-Verse FID-1 41.9 #64 of 69 Archive leaderboard report
Text-to-Image Generation COCO (Common Objects in Context) L-Verse FID-2 35.5 #64 of 69 Archive leaderboard report
Text-to-Image Generation COCO (Common Objects in Context) L-Verse FID-4 30.2 #64 of 69 Archive leaderboard report
Text-to-Image Generation COCO (Common Objects in Context) L-Verse FID-8 29.83 #64 of 69 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.

Methods

AttentionBPELinear LayerMulti-Head AttentionResidual ConnectionTransformerVQ-VAE

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