update

app.py

@@ -13,18 +13,18 @@ import torch.nn.functional as F
 from torchvision.transforms import Compose
 import tempfile
 from gradio_imageslider import ImageSlider
-from .depth_anything.depth_anything.dpt import DepthAnything
-from .depth_anything.depth_anything.util.transform import Resize, NormalizeImage, PrepareForNet
+from depth_anything.depth_anything.dpt import DepthAnything
+from depth_anything.depth_anything.util.transform import Resize, NormalizeImage, PrepareForNet
 
 NUM_INFERENCE_STEPS = 50
 dtype = torch.float16
 if torch.cuda.is_available():
-  device = "cuda"
+  DEVICE = "cuda"
 elif torch.backends.mps.is_available():
-  device = "mps"
+  DEVICE = "mps"
   dtype = torch.float32
 else:
-  device = "cpu"
+  DEVICE = "cpu"
 #device = "cuda"
 
 encoder = 'vitl' # can also be 'vitb' or 'vitl'
@@ -92,7 +92,7 @@ def preprocess_image(image):
     image = transforms.CenterCrop((image.size[1] // 64 * 64, image.size[0] // 64 * 64))(image)
     image = transforms.ToTensor()(image)
     image = image * 2 - 1
-    image = image.unsqueeze(0).to(device)
+    image = image.unsqueeze(0).to(DEVICE)
     return image
 
 
@@ -101,7 +101,7 @@ def preprocess_map(map):
     map = transforms.CenterCrop((map.size[1] // 64 * 64, map.size[0] // 64 * 64))(map)
     # convert to tensor
     map = transforms.ToTensor()(map)
-    map = map.to(device)
+    map = map.to(DEVICE)
     return map
 
 
@@ -109,14 +109,14 @@ def inference(image, map, gs, prompt, negative_prompt):
     validate_inputs(image, map)
     image = preprocess_image(image)
     map = preprocess_map(map)
-    base_cuda = base.to(device)
+    base_cuda = base.to(DEVICE)
     edited_images = base_cuda(prompt=prompt, original_image=image, image=image, strength=1, guidance_scale=gs,
                          num_images_per_prompt=1,
                          negative_prompt=negative_prompt,
                          map=map,
                          num_inference_steps=NUM_INFERENCE_STEPS, denoising_end=0.8, output_type="latent").images
     base_cuda=None
-    refiner_cuda = refiner.to(device)
+    refiner_cuda = refiner.to(DEVICE)
     edited_images = refiner_cuda(prompt=prompt, original_image=image, image=edited_images, strength=1, guidance_scale=7.5,
                             num_images_per_prompt=1,
                             negative_prompt=negative_prompt,
@@ -144,20 +144,21 @@ with gr.Blocks() as demo:
         with gr.Column():
             with gr.Row():
                 input_image = gr.Image(label="Input Image", type="pil")
-                change_map = gr.Image(label="Change Map", type="pil")
+#                change_map = gr.Image(label="Change Map", type="pil")
             gs = gr.Slider(0, 28, value=7.5, label="Guidance Scale")
             prompt = gr.Textbox(label="Prompt")
             neg_prompt = gr.Textbox(label="Negative Prompt")
             with gr.Row():
-                clr_btn=gr.ClearButton(components=[input_image, change_map, gs, prompt, neg_prompt])
+#                clr_btn=gr.ClearButton(components=[input_image, change_map, gs, prompt, neg_prompt])
+                clr_btn=gr.ClearButton(components=[input_image, gs, prompt, neg_prompt])
                 run_btn = gr.Button("Run",variant="primary")
 
         output = gr.Image(label="Output Image")
     run_btn.click(
       run,
       #inference,
-      inputs=[input_image, change_map, gs, prompt, neg_prompt],
-      outputs=output
+      inputs=[input_image, gs, prompt, neg_prompt],
+      outputs=[change_map, output]
     )
     clr_btn.add(output)
 if __name__ == "__main__":