Video Motion Transfer with Diffusion Transformers
Paper
•
2412.07776
•
Published
•
16
Davide Buoso
lambdavi
·
AI & ML interests
Data Science student @ Polytechnic University of Turin.
Interested in the intersection of Robotics and Reinforcement Learning.
Recent Activity
upvoted
a
paper
10 days ago
Video Motion Transfer with Diffusion Transformers
liked
a dataset
15 days ago
kelvin34501/OakInk-v2
updated
a model
16 days ago
lambdavi/span-marker-luke-legal
Organizations
None yet
lambdavi's activity
upvoted
a
paper
10 days ago
liked
a
dataset
15 days ago
updated
a
model
16 days ago
upvoted
a
paper
16 days ago
reacted to
akhaliq's
post with 👍
7 months ago
Post
20881
Chameleon
Mixed-Modal Early-Fusion Foundation Models
Chameleon: Mixed-Modal Early-Fusion Foundation Models (2405.09818)
We present Chameleon, a family of early-fusion token-based mixed-modal models capable of understanding and generating images and text in any arbitrary sequence. We outline a stable training approach from inception, an alignment recipe, and an architectural parameterization tailored for the early-fusion, token-based, mixed-modal setting. The models are evaluated on a comprehensive range of tasks, including visual question answering, image captioning, text generation, image generation, and long-form mixed modal generation. Chameleon demonstrates broad and general capabilities, including state-of-the-art performance in image captioning tasks, outperforms Llama-2 in text-only tasks while being competitive with models such as Mixtral 8x7B and Gemini-Pro, and performs non-trivial image generation, all in a single model. It also matches or exceeds the performance of much larger models, including Gemini Pro and GPT-4V, according to human judgments on a new long-form mixed-modal generation evaluation, where either the prompt or outputs contain mixed sequences of both images and text. Chameleon marks a significant step forward in a unified modeling of full multimodal documents.
Mixed-Modal Early-Fusion Foundation Models
Chameleon: Mixed-Modal Early-Fusion Foundation Models (2405.09818)
We present Chameleon, a family of early-fusion token-based mixed-modal models capable of understanding and generating images and text in any arbitrary sequence. We outline a stable training approach from inception, an alignment recipe, and an architectural parameterization tailored for the early-fusion, token-based, mixed-modal setting. The models are evaluated on a comprehensive range of tasks, including visual question answering, image captioning, text generation, image generation, and long-form mixed modal generation. Chameleon demonstrates broad and general capabilities, including state-of-the-art performance in image captioning tasks, outperforms Llama-2 in text-only tasks while being competitive with models such as Mixtral 8x7B and Gemini-Pro, and performs non-trivial image generation, all in a single model. It also matches or exceeds the performance of much larger models, including Gemini Pro and GPT-4V, according to human judgments on a new long-form mixed-modal generation evaluation, where either the prompt or outputs contain mixed sequences of both images and text. Chameleon marks a significant step forward in a unified modeling of full multimodal documents.
liked
2
models
9 months ago
updated
a
model
10 months ago
updated
a
collection
10 months ago
updated
3
models
10 months ago
updated
a
collection
11 months ago
updated
a
model
11 months ago
liked
a
Space
11 months ago
updated
a
model
11 months ago
updated
a
collection
12 months ago