DeepSeek Janus-Pro: DeepSeek's Revolution in Multimodal AI?
In this video we dive into DeepSeek's Janus-Pro, a new multimodal AI model, released by DeepSeek not long after the groundbreaking release of DeepSeek-R1. Can Janus-Pro create a similar revolution, this time in multimodal AI?
Janus-Pro was introduced in a paper titled, "Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling".
This paper builds on a previous paper by DeepSeek, titled "JanusFlow: Harmonizing Autoregression and Rectified Flow for Unified Multimodal Understanding and Generation".
To understand this release, we'll explain both papers in this video.
Written Review - https://aipapersacademy.com/janus-pro/
Janus-Pro paper - https://github.com/deepseek-ai/Janus/...
Janus paper - https://arxiv.org/abs/2411.07975
-----------------------------------------------------------------------------------------------
✉️ Join the newsletter - https://aipapersacademy.com/newsletter/
👍 Please like & subscribe if you enjoy this content
Become a patron - / aipapersacademy
The video was edited using VideoScribe - https://tidd.ly/44TZEiX
-----------------------------------------------------------------------------------------------
Chapters:
0:00 Introduction
1:00 Understanding & Generation Tasks
2:32 Janus - A Unified Model
3:09 Janus-Pro Architecture
5:43 Training Process
7:38 Janus-Pro Results