Vision and Language Pre-Trained Models
OSCAR
Introduced by Li et al. in Oscar: Object-Semantics Aligned Pre-training for Vision-Language TasksOSCAR is a new learning method that uses object tags detected in images as anchor points to ease the learning of image-text alignment. The model take a triple as input (word-tag-region) and pre-trained with two losses (masked token loss over words and tags, and a contrastive loss between tags and others). OSCAR represents an image-text pair into semantic space via dictionary lookup. Object tags are used as anchor points to align image regions with word embeddings of pre-trained language models. The model is then fine-tuned for understanding and generation tasks.
Source: Oscar: Object-Semantics Aligned Pre-training for Vision-Language Tasks
Papers
| Paper | Code | Results | Date | Stars |
|---|
Tasks
| Task | Papers | Share |
|---|---|---|
| Language Modelling | 4 | 11.43% |
| Image Captioning | 3 | 8.57% |
| Visual Question Answering (VQA) | 3 | 8.57% |
| Question Answering | 2 | 5.71% |
| NER | 2 | 5.71% |
| Change Detection | 1 | 2.86% |
| GPT-4 | 1 | 2.86% |
| Image Generation | 1 | 2.86% |
| Text-to-Image Generation | 1 | 2.86% |
Usage Over Time
Components
| Component | Type |
|
|---|---|---|
| 🤖 No Components Found | You can add them if they exist; e.g. Mask R-CNN uses RoIAlign |
