ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks

Code

• github.com/hwanheelee1993/vilbertscore
  pytorch★ 18
• github.com/zihaow123/unimm
  pytorch★ 13
• github.com/vmurahari3/visdial-bert
  pytorch★ 0
• github.com/jiasenlu/vilbert_beta
  pytorch★ 0
• github.com/facebookresearch/vilbert-multi-task
  pytorch★ 0
• github.com/johntiger1/multitask_multimodal
  pytorch★ 0
• github.com/Mehrab-Tanjim/vilbert-rationalization
  pytorch★ 0
• github.com/fuqianya/ViLBERT-Paddle
  paddle★ 0
• github.com/jialinwu17/tmpimgs
  pytorch★ 0
• github.com/allenai/allennlp-models
  pytorch★ 0

Abstract

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.

Tasks

Benchmark Results

Reproductions