Open raghavmittal101 opened 3 months ago
Hi, my 2D edit result is awful. Could you provide more details about the the Painting-to-Real section? like the training images and their prompt, or the parameters etc.
This was the result I got from a Monet painting I found from the internet: There is just a correction to the code you need to add involving the src/tgt prompts (then for base prompt just type in a caption and for extra src prompt just add painting while leaving extra tgt blank):
Base prompt: A serene garden scene with a gently arched wooden bridge spanning across a pond filled with blooming water lilies.
loss_dict = guidance.bridge_stage_two(im=im, prompt=args.prompt, cfg_scale=args.cfg_scale, return_dict=True, extra_src_prompts=args.extra_src_prompt, extra_tgt_prompts=args.extra_tgt_prompt)
What is the method for giving an artistic/painting image as an input to output a realistic image? In the readme, I see there is a method for text-based NeRF optimization only.