VL-BERT: Pre-training of Generic Visual-Linguistic Representations

Eighth International Conference on Learning Representations (ICLR) |

PDF

We introduce a new pre-trainable generic representation for visual-linguistic tasks, called Visual-Linguistic BERT (VL-BERT for short). VL-BERT adopts the simple yet powerful Transformer model as the backbone, and extends it to take both visual and linguistic embedded features as input. In it, each element of the input is either of a word from the input sentence, or a region-of-interest (RoI) from the input image. It is designed to fit for most of the visual-linguistic downstream tasks. To better exploit the generic representation, we pre-train VL-BERT on the massive-scale Conceptual Captions dataset, together with text-only corpus. Extensive empirical analysis demonstrates that the pre-training procedure can better align the visual-linguistic clues and benefit the downstream tasks, such as visual commonsense reasoning, visual question answering and referring expression comprehension. It is worth noting that VL-BERT achieved the first place of single model on the leaderboard of the VCR benchmark.

Publication Downloads

VL-BERT

April 17, 2020

VL-BERT is a simple yet powerful pre-trainable generic representation for visual-linguistic tasks. It is pre-trained on the massive-scale caption dataset and text-only corpus, and can be fine-tuned for various down-stream visual-linguistic tasks, such as Visual Commonsense Reasoning, Visual Question Answering and Referring Expression Comprehension.