Retrieval-Guided and Semantically Grounded Image Captioning for Open-Domain Scenes
Recent image captioning methods based on pre-trained vision–language models can generate fluent and coherent descriptions, yet they still struggle in open-domain scenes that contain long-tail concepts, uncommon object combinations, and ambiguous visual evidence. Two limitations are especially import...
I tiakina i:
| Ngā kaituhi matua: | , , , |
|---|---|
| Hōputu: | Artigo |
| Reo: | Inglês |
| I whakaputaina: |
MDPI AG
2026-05-01
|
| Rangatū: | Mathematics |
| Ngā marau: | |
| Urunga tuihono: | https://www.mdpi.com/2227-7390/14/10/1667 |
| Ngā Tūtohu: |
Kāore He Tūtohu, Me noho koe te mea tuatahi ki te tūtohu i tēnei pūkete!
|