What Vision-Language Models `See' when they See Scenes
arXiv:2109.07301
Abstract
Images can be described in terms of the objects they contain, or in terms of the types of scene or place that they instantiate. In this paper we address to what extent pretrained Vision and Language models can learn to align descriptions of both types with images. We compare 3 state-of-the-art models, VisualBERT, LXMERT and CLIP. We find that (i) V&L models are susceptible to stylistic biases acquired during pretraining; (ii) only CLIP performs consistently well on both object- and scene-level descriptions. A follow-up ablation study shows that CLIP uses object-level information in the visual modality to align with scene-level textual descriptions.
References in corpus (8)
- Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
- Language Models are Few-Shot Learners
- Microsoft COCO Captions: Data Collection and Evaluation Server
- VisualBERT: A Simple and Performant Baseline for Vision and Language
- ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision
- MiniVLM: A Smaller and Faster Vision-Language Model
- Are Convolutional Neural Networks or Transformers more like human vision?
- SemVLP: Vision-Language Pre-training by Aligning Semantics at Multiple Levels