add deocemntation, imporve run time on inversion

.gitattributes

@@ -29,6 +29,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.tgz filter=lfs diff=lfs merge=lfs -text
 *.wasm filter=lfs diff=lfs merge=lfs -text
 *.xz filter=lfs diff=lfs merge=lfs -text
+*.dat filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text

app.py

@@ -10,7 +10,11 @@ def greet(image,text,power):
     image_edit = model_helper.image_from_text(text,PIL_image,power)
     return image_edit
 
-description = "demo for model to edit face with text, you can see the [github repo CelebrityLook](https://github.com/amit154154/CelebrityLook)"
+description = "demo for model to edit face with text, you can see the [github repo CelebrityLook](https://github.com/amit154154/CelebrityLook) \n" \
+              "note that the model is in alpha version, so it's not perfect, but it's fun to play with it, some guidelines:\n " \
+              "1. the image you give the model need to be aligin to the camera such that your eyes are infront of the camera\n" \
+              "2.play with the power of the text and the text itself, it will look like 💩 sometimes" \
+
 title = "FaceOver - edit face with text 🐨 "
 
 examples = [

help_function.py

@@ -4,6 +4,9 @@ from torchvision import transforms
 from torchvision.transforms import ToPILImage
 import torch.nn.functional as F
 
+
+
+
 class help_function:
     def __init__(self):
         self.clip_text_model = torch.jit.load('jit_models/clip_text_jit.pt', map_location=torch.device('cpu'))
@@ -36,11 +39,15 @@ class help_function:
         w_delta = self.mapper_clip(text_feachers - self.mean_clip)
         return w_delta
 
+
     def image_from_text(self,text,image,power = 1.0):
         w_inversion = self.get_image_inversion(image)
-        text_embedding = self.get_text_embedding(text)
-        w_delta = self.get_text_delta(text_embedding)
-
-        w_edit = w_inversion + w_delta * power
+        if power != 0:
+            text_embedding = self.get_text_embedding(text)
+            w_delta = self.get_text_delta(text_embedding)
+            w_edit = w_inversion + w_delta * power
+        else:
+            w_edit = w_inversion
         image_edit = self.decoder(w_edit)
-        return ToPILImage()((image_edit[0]+0.5)*0.5)
+        image_edit = ToPILImage()((image_edit[0]+0.5)*0.5).resize((512,512))
+        return image_edit