add image

.gitignore

@@ -3,3 +3,4 @@ results_100/
 prompt_to_prompt/__pycache__/utils.cpython-38.pyc
 prompt_to_prompt/__pycache__/*
 *.pyc
+results/

README.md

@@ -59,16 +59,18 @@ These three inputs are supplied through a YAML file (please use the provided 'te
 Method's parameters are:
 
 ```
-skip - controlling the adherence to the input image
-cfg_tar - the strength of the classifier free guidance
+skip - controlling the adherence to the input image; a list
+cfg_tar - classifier free guidance strengths; a list
 ```
 These two parameters have default values, as descibed in the paper.
 
 ## Usage Example 
 ```
-python3 main_run.py --mode="XXX" --dataset_yaml="test.yaml" --skip=36 --cfg_tar=15 
+python3 main_run.py --mode="our_inv" --dataset_yaml="test.yaml" --skip=[36] --cfg_tar=[15] 
+python3 main_run.py --mode="p2pinv" --dataset_yaml="test.yaml" --skip=[12] --cfg_tar=[9] 
+
 ```
-Where XXX can be ```our_inv```,```p2pinv``` ,```ddim``` or ```p2p```.
+The mode argument can also be: ```ddim``` or ```p2p```.
 
 In ```our_inv``` and ```p2pinv``` modes we suggest to play around with ```skip``` in the range [0,40] and ```cfg_tar``` in the range [7,18].
 

ddm_inversion/ddim_inversion.py

@@ -1,5 +1,5 @@
 
-from prompt_to_prompt.inversion_utils import encode_text
+from ddm_inversion.inversion_utils import encode_text
 from typing import Union
 import torch
 import numpy as np

ddm_inversion/inversion_utils.py

@@ -3,7 +3,7 @@
 from tqdm import tqdm
 
 def load_real_image(folder = "data/", img_name = None, idx = 0, img_size=512, device='cuda'):
-    from prompt_to_prompt.utils import pil_to_tensor
+    from ddm_inversion.utils import pil_to_tensor
     from PIL import Image
     from glob import glob
     if img_name is not None:

main_run.py

@@ -15,14 +15,14 @@
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()
-    parser.add_argument("--device_num", type=int, default=0)
+    parser.add_argument("--device_num", type=int, default=2)
     parser.add_argument("--cfg_src", type=float, default=3.5)
     parser.add_argument("--cfg_tar", type=float, default=[15])
     parser.add_argument("--num_diffusion_steps", type=int, default=100)
     parser.add_argument("--dataset_yaml",  default="test.yaml")
     parser.add_argument("--eta", type=float, default=1)
-    parser.add_argument("--mode",  default="our_inv", help="modes: our_inv,p2pinv,p2pddim, ddim")
-    parser.add_argument("--skip",  type=int, default=[0])
+    parser.add_argument("--mode",  default="our_inv", help="modes: our_inv,p2pinv,p2pddim,ddim")
+    parser.add_argument("--skip",  type=int, default=[36])
     parser.add_argument("--xa", type=float, default=0.6)
     parser.add_argument("--sa", type=float, default=0.2)
 
@@ -146,4 +146,7 @@
                     save_full_path = os.path.join(save_path, image_name_png)
                     img.save(save_full_path)
 
-
+# TODO: Inbar:
+# (1) fix Rene's bug
+# (2) write more concisly the inversion_util
+# (3) Comments

test.yaml

@@ -1,6 +1,15 @@
+  # -
+  #   init_img:  /example_images/horse_mud.jpg
+  #   source_prompt: a photo of a horse in the mud
+
+  #   target_prompts:
+  #   - a photo of a horse in the snow 
+  #   - a photo of a zebra in the snow 
+  #   - a photo of a zebra in the mud 
+
   -
-    init_img:  /example_images/horse_mud.jpg
-    source_prompt: a photo of a horse in the mud
+    init_img:  /example_images/sketch_cat.jpg
+    source_prompt: a sketch of a cat
 
     target_prompts:
-    - a photo of a horse in the snow 
+    - a sculpture of a cat