Diffusion Models Websites
DS-Fusion: Artistic Typography via Discriminated and Stylized Diffusio
DS-Fusion: create artistic typography automatically
FABRIC 🎨
Personalizing Diffusion Models with Iterative Feedback
Arijit Ray's Webpage
I am a Computer Vision Ph.D. Student at Boston University. I am excited about how to make human-AI and AI-AI teams solve tasks effectively. Consequently, I am interested in and work on models that can interact with humans using natural language, models that can rationalize and explain their decisions, and models that l
Symbolic Music Generation with Non-Differentiable Rule Guided Diffusio
Symbolic Music Generation with Non-Differentiable Rule Guided Diffusion.
3D Diffuser Actor: Policy Diffusion with 3D Scene Representations
ConceptBed
Large-scale concept learning evaluations for personalized text-to-image diffusion models.
Generative Models: What do they know?
Generative Models capture intrinsic image representations and we show how to extract them using a unified approach that works for any generative model including autoregressive, gan and difussion.
Diffuser: Reinforcement Learning with Diffusion Models
Diffusion models for reinforcement learning and planning
ReGround: Improving Textual and Spatial Grounding at No Cost
ReGround: Improving Textual and Spatial Grounding at No Cost
Differential Diffusion: Giving Each Pixel Its Strength
Editing different parts of a picture by varying amounts, as specified by a map
Fairy: Fast Parallelized Instruction-Guided Video-to-Video Synthesis
Project webpage for fairy-video2video
Diffusion Classifier
Diffusion Classifier leverages pretrained diffusion models to perform zero-shot classification without additional training.
Posterior Distillation Sampling
text-driven parametric image editing that matches the latents encoded in the posterior of a diffusion model’s forward process.
SyncDiffusion: Coherent Montage via Synchronized Joint Diffusions
SyncDiffusion synchronizes multiple diffusions to create coherent image montages.
Diffusion Hyperfeatures: Searching Through Time and Space for Semantic
Diffusion Hyperfeatures is a framework for consolidating the multi-scale and multi-timestep internal representations of a diffusion model for tasks such as semantic correspondence.
Diffusion-TTA
Generative diffusion models are great test-time adapters for discriminative models.
Align-Prop
We propose AlignProp, a method that uses reward backpropogation for the alignment of large-scale text-to-image diffusion models.
LfVoid: Can Pre-Trained Text-to-Image Models Generate Visual Goals for
Learning table-top manipulation tasks using goal images generated by pre-trained text-to-image models