LaDiC: Are Diffusion Models Really Inferior to Autoregressive Counterparts for Image-to-Text Generation?
Yuchi WangShuhuai RenRundong GaoLinli YaoQingyan GuoKaikai AnJianhong BaiXu SunPublished in: NAACL-HLT (2024)
Keyphrases
- autoregressive
- gaussian markov random field
- random fields
- image data
- input image
- multiscale
- diffusion models
- image analysis
- non stationary
- image segmentation
- edge detection
- graphical models
- high resolution
- natural language generation
- text generation
- segmentation method
- multiresolution
- information diffusion
- bayesian networks
- machine learning