Microsoft's open source Visual ChatGPT has won 20000 stars in five days

Source: contribution
Author: Rocky
2023-03-15 09:10:00

In addition to investing heavily in Open AI, Microsoft also stepped down to make a big deal in AI. Five days ago, Microsoft opened the source Visual ChatGPT This software can connect ChatGPT and a series of visual models to achieve Sending and receiving images

As we all know, although ChatGPT is very powerful and can even be used to write novels and papers, it is currently limited to text communication. But the emoticon pack has already become an indispensable function of daily text chat.

The appearance of Visual ChatGPT is like the first addition of the emoticon package function to the text communication APP, and it is also a "customized emoticon package" automatically generated according to the text input by users, which greatly improves the interest and application field of ChatGPT.

On the one hand, ChatGPT (or LLM) serves as a general interface, providing image understanding and user interaction functions. On the other hand, the basic image model serves as the technical expert behind it by providing in-depth knowledge in specific fields.

The technical architecture and schematic diagram are listed in the warehouse:

There are three different types of conversations in Demo, namely Visual ChatGPT receive The user's image and Visual ChatGPT are based on the user's text Modify image and send To users and Visual ChatGPT Identify pictures , and answer users' questions. Visual ChatGPT will judge whether to use the VFM (Visual Foundation Model) This problem.

The warehouse also shows the image model and video storage usage used by Visual ChatGPT:

For more details, you can read the arxiv paper of Visual ChatGPT: Visual ChatGPT: Talking, Drawing and Editing with Visual Foundation Models

Visual ChatGPT was released on March 10. As of 9:00 a.m. on March 15, the project has temporarily received 19547 Stars, which can be described as a rocket rise.

Expand to read the full text
Click to join the discussion 🔥 (18) Post and join the discussion 🔥
This wonderful review
It is predicted that Wenxin will attack the street in a word, and who has confidence in Baidu in what era
2023-03-16 09:27
two fabulous
report
In this prediction, classical Chinese is all about music
2023-03-15 16:50
one fabulous
report
These two days are full of heavy bombs, chatGPT4.0,visualGPT, Tomorrow, I will wait for the classical Chinese to amaze me..
2023-03-15 10:09
one fabulous
report
eighteen comment
sixteen Collection
 Back to top
Top