swMATH ID: 42498
Software Authors: Jiasen Lu, Dhruv Batra, Devi Parikh, Stefan Lee
Description: ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks. We present ViLBERT (short for Vision-and-Language BERT), a model for learning task-agnostic joint representations of image content and natural language. We extend the popular BERT architecture to a multi-modal two-stream model, pro-cessing both visual and textual inputs in separate streams that interact through co-attentional transformer layers. We pretrain our model through two proxy tasks on the large, automatically collected Conceptual Captions dataset and then transfer it to multiple established vision-and-language tasks – visual question answering, visual commonsense reasoning, referring expressions, and caption-based image retrieval – by making only minor additions to the base architecture. We observe significant improvements across tasks compared to existing task-specific models – achieving state-of-the-art on all four tasks. Our work represents a shift away from learning groundings between vision and language only as part of task training and towards treating visual grounding as a pretrainable and transferable capability.
Homepage: https://arxiv.org/abs/1908.02265
Source Code:  https://github.com/facebookresearch/vilbert-multi-task
Related Software: Adam; VideoBERT; Flickr30K; ImageNet; BLEU; S4L; VQA; BERT; Rouge; GloVe; LXMERT; VisualBERT; PointNet; MNIST; CamNet; MVSNet; Fashion-MNIST; SynSin; Make3D; PIFuHD
Cited in: 2 Publications

Citations by Year