Scene Graph Generation via Conditional Random Fields
Fuente:
arXiv
Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2018
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917572874600448 |
|---|---|
| author | Cong, Weilin Wang, William Lee, Wang-Chien |
| author_facet | Cong, Weilin Wang, William Lee, Wang-Chien |
| contents | Despite the great success object detection and segmentation models have achieved in recognizing individual objects in images, performance on cognitive tasks such as image caption, semantic image retrieval, and visual QA is far from satisfactory. To achieve better performance on these cognitive tasks, merely recognizing individual object instances is insufficient. Instead, the interactions between object instances need to be captured in order to facilitate reasoning and understanding of the visual scenes in an image. Scene graph, a graph representation of images that captures object instances and their relationships, offers a comprehensive understanding of an image. However, existing techniques on scene graph generation fail to distinguish subjects and objects in the visual scenes of images and thus do not perform well with real-world datasets where exist ambiguous object instances. In this work, we propose a novel scene graph generation model for predicting object instances and its corresponding relationships in an image. Our model, SG-CRF, learns the sequential order of subject and object in a relationship triplet, and the semantic compatibility of object instance nodes and relationship nodes in a scene graph efficiently. Experiments empirically show that SG-CRF outperforms the state-of-the-art methods, on three different datasets, i.e., CLEVR, VRD, and Visual Genome, raising the Recall@100 from 24.99% to 49.95%, from 41.92% to 50.47%, and from 54.69% to 54.77%, respectively. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_1811_08075 |
| institution | arXiv |
| publishDate | 2018 |
| record_format | arxiv |
| spellingShingle | Scene Graph Generation via Conditional Random Fields Cong, Weilin Wang, William Lee, Wang-Chien Computer Vision and Pattern Recognition Despite the great success object detection and segmentation models have achieved in recognizing individual objects in images, performance on cognitive tasks such as image caption, semantic image retrieval, and visual QA is far from satisfactory. To achieve better performance on these cognitive tasks, merely recognizing individual object instances is insufficient. Instead, the interactions between object instances need to be captured in order to facilitate reasoning and understanding of the visual scenes in an image. Scene graph, a graph representation of images that captures object instances and their relationships, offers a comprehensive understanding of an image. However, existing techniques on scene graph generation fail to distinguish subjects and objects in the visual scenes of images and thus do not perform well with real-world datasets where exist ambiguous object instances. In this work, we propose a novel scene graph generation model for predicting object instances and its corresponding relationships in an image. Our model, SG-CRF, learns the sequential order of subject and object in a relationship triplet, and the semantic compatibility of object instance nodes and relationship nodes in a scene graph efficiently. Experiments empirically show that SG-CRF outperforms the state-of-the-art methods, on three different datasets, i.e., CLEVR, VRD, and Visual Genome, raising the Recall@100 from 24.99% to 49.95%, from 41.92% to 50.47%, and from 54.69% to 54.77%, respectively. |
| title | Scene Graph Generation via Conditional Random Fields |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/1811.08075 |