 Fully Convolutional Scene Graph Generation/Pasted_image_20250314123239.png)
 Fully Convolutional Scene Graph Generation/Pasted_image_20250317115153.png)
 Fully Convolutional Scene Graph Generation/Pasted_image_20250317115203.png)
 Fully Convolutional Scene Graph Generation/Pasted_image_20250317115216.png)
 Fully Convolutional Scene Graph Generation/Pasted_image_20250317121116.png)
 Fully Convolutional Scene Graph Generation/Pasted_image_20250317121142.png)
 Fully Convolutional Scene Graph Generation/Pasted_image_20250317121325.png)
Energy-Based Learning for Scene Graph Generation
Image Retrieval using Scene Graphs
Building an efficient structured representation that captures comprehensive semantic knowledge is a crucial step towards a deeper understanding of visual scenes
SGTR+= End-to-end Scene Graph Generation with Transformer
SGTR 是一种自上而下的方法，该方法首先使用基于Transformer的生成器来生成一组可学习的triplet queries (subject–predicate–object)，然后使用级联的triplet detector逐步完善这些查询并生成最终场景图。它还提出了一种基于结构化发生器的实体感知关系表示方法，该方法利用了关系的组成属性。
Top-down approach (SGTR):
- Starts with higher-level structures (triplet queries) and refines them
- Begins by generating complete subject-predicate-object triplet candidates
- Then progressively refines these triplets to match the image content
- Works with the complete structural units from the beginning
- Analogous to starting with a rough sketch of the entire tree and then refining each branch
