moondream
moondream copied to clipboard
Larger images for detail recognition?
The current model is working only with 378x378 image resolution. Is it possible to make it recognize images with higher resolution to extract more details?
Definitely in the roadmap!
@vikhyat I’m glad to hear that! Also, I’m wondering about the training. Can you please provide me with a little more details about how I can train the model? I saw there’s a code, but do I need to write some captions, downscale images, some masks, or whatever?