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

NeurIPS 2019  ·  Jiasen Lu, Dhruv Batra, Devi Parikh, Stefan Lee ·

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.

PDF Abstract NeurIPS 2019 PDF NeurIPS 2019 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Visual Question Answering (VQA) A-OKVQA ViLBERT - OK-VQA MC Accuracy 34.1 # 9
DA VQA Score 9.2 # 10
Visual Question Answering (VQA) A-OKVQA ViLBERT - VQA MC Accuracy 42.1 # 5
DA VQA Score 12.0 # 9
Visual Question Answering (VQA) A-OKVQA ViLBERT MC Accuracy 41.5 # 7
DA VQA Score 25.9 # 6
Referring Expression Comprehension Talk2Car Vilbert (Base) AP50 68.9 # 5
Visual Question Answering (VQA) VQA v2 test-dev ViLBERT Accuracy 70.55 # 29