AI Image Analyzer
This is a simple AI Image Analyzer that uses the ollama,mistral.rs API or transformers to analyze images. The user can upload an image and the AI will analyze the image and provide a description of the image. The user can also translate the description to a different language.
The original intention of the project is to make a small visual model tool for doing some tests or some simple applications, such as picture description, picture translation, etc. It may not be perfect but it's a starting point for you to start understanding visual models.
If it is helpful to you, I hope you will not hesitate to give me a star. This will be a recognition of my work. If you have interesting ideas, you can submit them to me on issues.
The current test found that only Molmo-7B-D-0924 will accurately return coordinate-related information(<point x="4.5" y="73.0" alt="电饭煲">电饭煲</point>). Other models seem to be unable to do so. I am trying to let minicpm return spatial information. I believe it will be updated soon.
Since the startup of loading the translation model will be slow, this is normal. Just wait for the web site(http://127.0.0.1:7777) to load. Don't worry.
- allenai/Molmo-7B-D-0924
- Helsinki-NLP/opus-mt-en-zh
