bug fix

app.py

@@ -2,9 +2,8 @@ import spaces
 import gradio as gr
 import numpy as np
 import random
-from diffusers import DiffusionPipeline
+import PIL.Image
 import torch
-import random
 from diffusers import (
     ControlNetModel,
     DiffusionPipeline,
@@ -62,10 +61,10 @@ def get_depth_map(image):
     image = feature_extractor(images=image, return_tensors="pt").pixel_values.to("cuda")
     with torch.no_grad(), torch.autocast("cuda"):
         depth_map = depth_estimator(image).predicted_depth
-
+    size = (image.shape[-2], image.shape[-1])
     depth_map = torch.nn.functional.interpolate(
         depth_map.unsqueeze(1),
-        size=(1024, 1024),
+        size=size,
         mode="bicubic",
         align_corners=False,
     )
@@ -80,20 +79,21 @@ def get_depth_map(image):
 
 
 @spaces.GPU(enable_queue=True)
-def process(orginal_image, image_url, prompt, n_prompt, num_steps, guidance_scale, control_strength, seed):
+def process(image, image_url, prompt, n_prompt, num_steps, guidance_scale, control_strength, seed):
 
     if image_url:
         orginal_image = load_image(image_url)
-    
-    width = 1024
-    height = 1024
-    depth_image = get_depth_map(orginal_image.resize((1024, 1024)))
+    else:
+        orginal_image = PIL.Image.fromarray(image)
+
+    size = (orginal_image.size[0], orginal_image.size[1])
+    depth_image = get_depth_map(orginal_image)
     generator = torch.Generator().manual_seed(seed)
-    generated_image = self.pipe(
+    generated_image = pipe(
         prompt=prompt,
         negative_prompt=n_prompt,
-        width=width,
-        height=height,
+        width=size[0],
+        height=size[1],
         guidance_scale=guidance_scale,
         num_inference_steps=num_steps,
         strength=control_strength,