Upload 2 files

src/rep_api.py

@@ -1,315 +1,380 @@
-import replicate 
-import os
-from src.utils import image_to_base64 , update_model_dicts, BB_uploadfile,numpy_to_base64
-from src.deepl import detect_and_translate
-import json
-import time
-style_json="model_dict.json"
-model_dict=json.load(open(style_json,"r"))
-
-
-def generate_image_control_net(prompt,lora_model,api_path,aspect_ratio,lora_scale,
-                              use_control_net,control_net_type,control_net_img,control_net_strength, 
-                             num_outputs=1,guidance_scale=3.5,seed=None,                
-): 
-   print(prompt,lora_model,api_path,aspect_ratio,use_control_net)
-   inputs = {
-       "prompt": detect_and_translate(prompt),
-        "output_format": "png",
-         "num_outputs":num_outputs,
-        "guidance_scale": guidance_scale,
-        "output_quality": 100,
-   }
-   if seed is not None:
-      inputs["seed"]=seed
-
-
-   if use_control_net:
-      api_path= "xlabs-ai/flux-dev-controlnet:f2c31c31d81278a91b2447a304dae654c64a5d5a70340fba811bb1cbd41019a2" #X labs control net replicate repo
-      lora_url=model_dict[lora_model][1]
-      assert control_net_img is not None, "Please add control net image"
-      control_net_img=image_to_base64(control_net_img)
-      inputs["lora_url"]=lora_url
-      inputs["prompt"]+=", "+model_dict[lora_model][2]
-      inputs["control_image"]=control_net_img
-      inputs["control_type"]=control_net_type
-      inputs["control_strengh"]=control_net_strength
-      #other settings 
-      inputs["steps"]=30
-      inputs["lora_strength"]=lora_scale
-      inputs["negative_prompt"]= "low quality, ugly, distorted, artefacts"
-   else:
-      api_path=model_dict[lora_model][0]
-      inputs["aspect_ratio"]=aspect_ratio
-      inputs["prompt"]+=", "+model_dict[lora_model][2]
-      inputs["num_inference_steps"]=28
-      inputs["model"]="dev"
-      inputs["lora_scale"]=lora_scale
-   
-   #run model
-   output = replicate.run(
-     api_path,
-     input=inputs
-   )
-   print(output)
-   return output[0]
-
-   
-
-
-def generate_image_replicate(prompt,lora_model,api_path,aspect_ratio,model,lora_scale,
-                             num_outputs=1,guidance_scale=3.5,seed=None,
-
-                             ):
-  print(prompt,lora_model,api_path,aspect_ratio)
-
-  #if model=="dev":
-  num_inference_steps=30
-  if model=="schnell":
-    num_inference_steps=5
-
-  if lora_model is not None:
-     api_path=model_dict[lora_model][0]
-
-  inputs={
-        "model": model,
-        "prompt": detect_and_translate(prompt),
-        "lora_scale":lora_scale,
-        "aspect_ratio": aspect_ratio,
-        "num_outputs":num_outputs,
-        "num_inference_steps":num_inference_steps,
-        "guidance_scale":guidance_scale,
-        "output_format":"png",
-    }
-  if seed is not None:
-    inputs["seed"]=seed
-  output = replicate.run(
-    api_path,
-    input=inputs
-  )
-  print(output)
-  return output[0]
-def replicate_bgcontrolnet(img,prompt,background_prompt, sampler_name= "DPM++ SDE Karras",
-        negative_prompt="(deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime, mutated hands and fingers:1.4), (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, disconnected limbs, mutation, mutated, ugly, disgusting, amputation"
-  ):
-    img=image_to_base64(img)
-    prompt=prompt+" ," +background_prompt
-    output=replicate.run(
-        "wolverinn/realistic-background:9f020c55e037529bf20ed1cb799d7aa290404cfbd45157686717ffc7ee511eab",
-    input={
-        "seed": -1,
-        "image":img,
-        "prompt":detect_and_translate(prompt),
-        "sampler_name":sampler_name,
-        "negative_prompt":negative_prompt
-    }
-    ) 
-    
-    return output["image"]
-
-def replicate_caption_api(image,model,context_text):
-  print(model,context_text)
-  base64_image = image_to_base64(image)
-  if model=="blip":
-    output = replicate.run(
-    "andreasjansson/blip-2:f677695e5e89f8b236e52ecd1d3f01beb44c34606419bcc19345e046d8f786f9",
-    input={
-        "image": base64_image,
-        "caption": True,
-        "question": context_text,
-        "temperature": 1,
-        "use_nucleus_sampling": False
-    }
-    )
-    print(output)
-
-  elif model=="llava-16":
-    output = replicate.run(
-       # "yorickvp/llava-13b:80537f9eead1a5bfa72d5ac6ea6414379be41d4d4f6679fd776e9535d1eb58bb",
-         "yorickvp/llava-v1.6-34b:41ecfbfb261e6c1adf3ad896c9066ca98346996d7c4045c5bc944a79d430f174",
-        input={
-            "image": base64_image,
-            "top_p": 1,
-            "prompt": context_text,
-            "max_tokens": 1024,
-            "temperature": 0.2
-          }
-        )
-    print(output)
-    output = "".join(output)
-
-  elif model=="img2prompt":
-    output = replicate.run(
-    "methexis-inc/img2prompt:50adaf2d3ad20a6f911a8a9e3ccf777b263b8596fbd2c8fc26e8888f8a0edbb5",
-    input={
-        "image":base64_image
-        }
-    )
-    print(output)
-  return output
-
-def update_replicate_api_key(api_key):
-    os.environ["REPLICATE_API_TOKEN"] = api_key
-    return f"Replicate API key updated: {api_key[:5]}..." if api_key else "Replicate API key cleared"
-
-
-def virtual_try_on(crop, seed, steps, category,  garm_img, human_img,  garment_des):
-    output = replicate.run(
-        "cuuupid/idm-vton:906425dbca90663ff5427624839572cc56ea7d380343d13e2a4c4b09d3f0c30f",
-        input={
-            "crop": crop,
-            "seed": seed,
-            "steps": steps,
-            "category": category,
-           # "force_dc": force_dc,
-            "garm_img": numpy_to_base64( garm_img),
-            "human_img":  numpy_to_base64(human_img),
-            #"mask_only": mask_only,
-            "garment_des": garment_des
-        }
-    )
-    print(output)
-    return output
-
-
-from src.utils import create_zip
-from PIL import Image
-
-
-def process_images(files,model,context_text,token_string):
-    images = []
-    textbox =""
-    for file in files:
-        print(file)
-        image = Image.open(file)
-        if model=="None":
-           caption="[Insert cap here]"
-        else:
-          caption = replicate_caption_api(image,model,context_text)
-        textbox += f"Tags: {caption}, file: " + os.path.basename(file) + "\n"
-        images.append(image)
-        #texts.append(textbox)
-    zip_path=create_zip(files,textbox,token_string)
-    
-    return images, textbox,zip_path
-
-def replicate_create_model(owner,name,visibility="private",hardware="gpu-a40-large"):
-  try:
-    model = replicate.models.create(
-        owner=owner,
-        name=name,
-        visibility=visibility,
-        hardware=hardware,
-    )
-    print(model)
-    return True
-  except Exception as e:
-    print(e)
-    if "A model with that name and owner already exists" in str(e):
-      return True
-    return False
-
-
-
-def traning_function(zip_path,training_model,training_destination,seed,token_string,max_train_steps,hf_repo_id=None,hf_token=None):
-  ##Place holder for now 
-  BB_bucket_name="jarvisdataset" 
-  BB_defult="https://f005.backblazeb2.com/file/"
-  if BB_defult not in zip_path:
-    zip_path=BB_uploadfile(zip_path,os.path.basename(zip_path),BB_bucket_name)
-  print(zip_path)
-  training_logs = f"Using zip traning file at: {zip_path}\n"
-  yield training_logs, None
-  input={
-              "steps": max_train_steps,
-              "lora_rank": 16,
-              "batch_size": 1,
-              "autocaption": True,
-              "trigger_word": token_string,
-              "learning_rate": 0.0004,
-              "seed": seed,
-              "input_images": zip_path
-          }
-  print(training_destination)
-  username,model_name=training_destination.split("/")
-  assert replicate_create_model(username,model_name,visibility="private",hardware="gpu-a40-large"),"Error in creating model on replicate, check API key and username is correct "
-
-  print(input)
-  try:
-      training = replicate.trainings.create(
-          destination=training_destination,
-          version="ostris/flux-dev-lora-trainer:1296f0ab2d695af5a1b5eeee6e8ec043145bef33f1675ce1a2cdb0f81ec43f02",
-          input=input,
-      )
-
-      training_logs = f"Training started with model: {training_model}\n"
-      training_logs += f"Destination: {training_destination}\n"
-      training_logs += f"Seed: {seed}\n"
-      training_logs += f"Token string: {token_string}\n"
-      training_logs += f"Max train steps: {max_train_steps}\n"
-
-      # Poll the training status
-      while training.status != "succeeded":
-          training.reload()
-          training_logs += f"Training status: {training.status}\n"
-          training_logs += f"{training.logs}\n"
-          if training.status == "failed":
-              training_logs += "Training failed!\n"
-              return training_logs, training
-              
-          yield training_logs, None
-          time.sleep(10)  # Wait for 10 seconds before checking again
-        
-      training_logs += "Training completed!\n"
-      if hf_repo_id and hf_token:
-          training_logs += f"Uploading to Hugging Face repo: {hf_repo_id}\n"
-          # Here you would implement the logic to upload to Hugging Face
-
-      traning_finnal=training.output
-
-      # In a real scenario, you might want to download and display some result images
-      # For now, we'll just return the original images
-      #images = [Image.open(file) for file in files]
-      _= update_model_dicts(traning_finnal["version"],token_string,style_json="model_dict.json")
-      traning_finnal["replicate_link"]="https://replicate.com/"+traning_finnal["version"].replace(":","/")
-      yield training_logs, traning_finnal
-
-  except Exception as e:
-        yield f"An error occurred: {str(e)}", None
-
-
-def sam_segment(image,prompt,negative_prompt,adjustment_factor=-15):
-  #img2 base64
-  image = image_to_base64(image)
-  output = replicate.run(
-      "schananas/grounded_sam:ee871c19efb1941f55f66a3d7d960428c8a5afcb77449547fe8e5a3ab9ebc21c",
-      input={
-          "image": image,
-          "mask_prompt": prompt,
-          "adjustment_factor": adjustment_factor,
-          "negative_mask_prompt":negative_prompt
-      }
-  )
-  out_items={}
-  for item in output:
-    # https://replicate.com/schananas/grounded_sam/api#output-schema
-    print(item)
-    out_items[os.path.basename(item).split(".")[0]]=item
-  return out_items 
-
-
-def replicate_zest(img,material_img="https://replicate.delivery/pbxt/Kl23gJODaW7EuxrDzBG9dcgqRdMaYSWmBQ9UexnwPiL7AnIr/3.jpg"):
-  if type(img)!=str:
-    img=image_to_base64(img)
-  if type(material_img)!=str:
-    material_img=image_to_base64(material_img)
-
-  output = replicate.run(
-      "camenduru/zest:11abc0a411459327938957581151c642dd1bee4cefe443a9a63b230c4fbc0952",
-      input={
-          "input_image": img,
-          "material_image":material_img
-      }
-  ) 
-  print(output)
-  return output
+import replicate 
+import os
+from src.utils import image_to_base64 ,open_image_from_url, update_model_dicts, BB_uploadfile,numpy_to_base64
+from src.deepl import detect_and_translate
+import json
+import time
+style_json="model_dict.json"
+model_dict=json.load(open(style_json,"r"))
+
+
+def generate_image_control_net(prompt,lora_model,api_path,aspect_ratio,lora_scale,
+                              use_control_net,control_net_type,control_net_img,control_net_strength, 
+                             num_outputs=1,guidance_scale=3.5,seed=None,                
+): 
+   print(prompt,lora_model,api_path,aspect_ratio,use_control_net)
+   inputs = {
+       "prompt": detect_and_translate(prompt),
+        "output_format": "png",
+         "num_outputs":num_outputs,
+        "guidance_scale": guidance_scale,
+        "output_quality": 100,
+   }
+   if seed is not None:
+      inputs["seed"]=seed
+
+
+   if use_control_net:
+      api_path= "xlabs-ai/flux-dev-controlnet:f2c31c31d81278a91b2447a304dae654c64a5d5a70340fba811bb1cbd41019a2" #X labs control net replicate repo
+      lora_url=model_dict[lora_model][1]
+      assert control_net_img is not None, "Please add control net image"
+      control_net_img=image_to_base64(control_net_img)
+      inputs["lora_url"]=lora_url
+      inputs["prompt"]+=", "+model_dict[lora_model][2]
+      inputs["control_image"]=control_net_img
+      inputs["control_type"]=control_net_type
+      inputs["control_strengh"]=control_net_strength
+      #other settings 
+      inputs["steps"]=30
+      inputs["lora_strength"]=lora_scale
+      inputs["negative_prompt"]= "low quality, ugly, distorted, artefacts"
+   else:
+      api_path=model_dict[lora_model][0]
+      inputs["aspect_ratio"]=aspect_ratio
+      inputs["prompt"]+=", "+model_dict[lora_model][2]
+      inputs["num_inference_steps"]=28
+      inputs["model"]="dev"
+      inputs["lora_scale"]=lora_scale
+   
+   #run model
+   output = replicate.run(
+     api_path,
+     input=inputs
+   )
+   print(output)
+   return output[0]
+
+   
+
+
+def generate_image_replicate(prompt,lora_model,api_path,aspect_ratio,model,lora_scale,
+                             num_outputs=1,guidance_scale=3.5,seed=None,
+
+                             ):
+  print(prompt,lora_model,api_path,aspect_ratio)
+
+  #if model=="dev":
+  num_inference_steps=30
+  if model=="schnell":
+    num_inference_steps=5
+
+  if lora_model is not None:
+     api_path=model_dict[lora_model][0]
+
+  inputs={
+        "model": model,
+        "prompt": detect_and_translate(prompt),
+        "lora_scale":lora_scale,
+        "aspect_ratio": aspect_ratio,
+        "num_outputs":num_outputs,
+        "num_inference_steps":num_inference_steps,
+        "guidance_scale":guidance_scale,
+        "output_format":"png",
+    }
+  if seed is not None:
+    inputs["seed"]=seed
+  output = replicate.run(
+    api_path,
+    input=inputs
+  )
+  print(output)
+  return output[0]
+def replicate_bgcontrolnet(img,prompt,background_prompt, sampler_name= "DPM++ SDE Karras",
+        negative_prompt="(deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime, mutated hands and fingers:1.4), (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, disconnected limbs, mutation, mutated, ugly, disgusting, amputation"
+  ):
+    img=image_to_base64(img)
+    prompt=prompt+" ," +background_prompt
+    output=replicate.run(
+        "wolverinn/realistic-background:9f020c55e037529bf20ed1cb799d7aa290404cfbd45157686717ffc7ee511eab",
+    input={
+        "seed": -1,
+        "image":img,
+        "prompt":prompt,
+        "sampler_name":sampler_name,
+        "negative_prompt":negative_prompt
+    }
+    ) 
+    
+    return output["image"]
+
+def replicate_caption_api(image,model,context_text):
+  print(model,context_text)
+  base64_image = image_to_base64(image)
+  if model=="blip":
+    output = replicate.run(
+    "andreasjansson/blip-2:f677695e5e89f8b236e52ecd1d3f01beb44c34606419bcc19345e046d8f786f9",
+    input={
+        "image": base64_image,
+        "caption": True,
+        "question": context_text,
+        "temperature": 1,
+        "use_nucleus_sampling": False
+    }
+    )
+    print(output)
+
+  elif model=="llava-16":
+    output = replicate.run(
+       # "yorickvp/llava-13b:80537f9eead1a5bfa72d5ac6ea6414379be41d4d4f6679fd776e9535d1eb58bb",
+         "yorickvp/llava-v1.6-34b:41ecfbfb261e6c1adf3ad896c9066ca98346996d7c4045c5bc944a79d430f174",
+        input={
+            "image": base64_image,
+            "top_p": 1,
+            "prompt": context_text,
+            "max_tokens": 1024,
+            "temperature": 0.2
+          }
+        )
+    print(output)
+    output = "".join(output)
+
+  elif model=="img2prompt":
+    output = replicate.run(
+    "methexis-inc/img2prompt:50adaf2d3ad20a6f911a8a9e3ccf777b263b8596fbd2c8fc26e8888f8a0edbb5",
+    input={
+        "image":base64_image
+        }
+    )
+    print(output)
+  return output
+
+def update_replicate_api_key(api_key):
+    os.environ["REPLICATE_API_TOKEN"] = api_key
+    return f"Replicate API key updated: {api_key[:5]}..." if api_key else "Replicate API key cleared"
+
+
+def virtual_try_on(crop, seed, steps, category,  garm_img, human_img,  garment_des):
+    output = replicate.run(
+        "cuuupid/idm-vton:906425dbca90663ff5427624839572cc56ea7d380343d13e2a4c4b09d3f0c30f",
+        input={
+            "crop": crop,
+            "seed": seed,
+            "steps": steps,
+            "category": category,
+           # "force_dc": force_dc,
+            "garm_img": numpy_to_base64( garm_img),
+            "human_img":  numpy_to_base64(human_img),
+            #"mask_only": mask_only,
+            "garment_des": garment_des
+        }
+    )
+    print(output)
+    return output
+
+
+from src.utils import create_zip
+from PIL import Image
+
+
+def process_images(files,model,context_text,token_string):
+    images = []
+    textbox =""
+    for file in files:
+        print(file)
+        image = Image.open(file)
+        if model=="None":
+           caption="[Insert cap here]"
+        else:
+          caption = replicate_caption_api(image,model,context_text)
+        textbox += f"Tags: {caption}, file: " + os.path.basename(file) + "\n"
+        images.append(image)
+        #texts.append(textbox)
+    zip_path=create_zip(files,textbox,token_string)
+    
+    return images, textbox,zip_path
+
+def replicate_create_model(owner,name,visibility="private",hardware="gpu-a40-large"):
+  try:
+    model = replicate.models.create(
+        owner=owner,
+        name=name,
+        visibility=visibility,
+        hardware=hardware,
+    )
+    print(model)
+    return True
+  except Exception as e:
+    print(e)
+    if "A model with that name and owner already exists" in str(e):
+      return True
+    return False
+
+
+
+def traning_function(zip_path,training_model,training_destination,seed,token_string,max_train_steps,hf_repo_id=None,hf_token=None):
+  ##Place holder for now 
+  BB_bucket_name="jarvisdataset" 
+  BB_defult="https://f005.backblazeb2.com/file/"
+  if BB_defult not in zip_path:
+    zip_path=BB_uploadfile(zip_path,os.path.basename(zip_path),BB_bucket_name)
+  print(zip_path)
+  training_logs = f"Using zip traning file at: {zip_path}\n"
+  yield training_logs, None
+  input={
+              "steps": max_train_steps,
+              "lora_rank": 16,
+              "batch_size": 1,
+              "autocaption": True,
+              "trigger_word": token_string,
+              "learning_rate": 0.0004,
+              "seed": seed,
+              "input_images": zip_path
+          }
+  print(training_destination)
+  username,model_name=training_destination.split("/")
+  assert replicate_create_model(username,model_name,visibility="private",hardware="gpu-a40-large"),"Error in creating model on replicate, check API key and username is correct "
+
+  print(input)
+  try:
+      training = replicate.trainings.create(
+          destination=training_destination,
+          version="ostris/flux-dev-lora-trainer:1296f0ab2d695af5a1b5eeee6e8ec043145bef33f1675ce1a2cdb0f81ec43f02",
+          input=input,
+      )
+
+      training_logs = f"Training started with model: {training_model}\n"
+      training_logs += f"Destination: {training_destination}\n"
+      training_logs += f"Seed: {seed}\n"
+      training_logs += f"Token string: {token_string}\n"
+      training_logs += f"Max train steps: {max_train_steps}\n"
+
+      # Poll the training status
+      while training.status != "succeeded":
+          training.reload()
+          training_logs += f"Training status: {training.status}\n"
+          training_logs += f"{training.logs}\n"
+          if training.status == "failed":
+              training_logs += "Training failed!\n"
+              return training_logs, training
+              
+          yield training_logs, None
+          time.sleep(10)  # Wait for 10 seconds before checking again
+        
+      training_logs += "Training completed!\n"
+      if hf_repo_id and hf_token:
+          training_logs += f"Uploading to Hugging Face repo: {hf_repo_id}\n"
+          # Here you would implement the logic to upload to Hugging Face
+
+      traning_finnal=training.output
+
+      # In a real scenario, you might want to download and display some result images
+      # For now, we'll just return the original images
+      #images = [Image.open(file) for file in files]
+      _= update_model_dicts(traning_finnal["version"],token_string,style_json="model_dict.json")
+      traning_finnal["replicate_link"]="https://replicate.com/"+traning_finnal["version"].replace(":","/")
+      yield training_logs, traning_finnal
+
+  except Exception as e:
+        yield f"An error occurred: {str(e)}", None
+
+
+def sam_segment(image,prompt,negative_prompt,adjustment_factor=-15):
+  #img2 base64
+  image = image_to_base64(image)
+  output = replicate.run(
+      "schananas/grounded_sam:ee871c19efb1941f55f66a3d7d960428c8a5afcb77449547fe8e5a3ab9ebc21c",
+      input={
+          "image": image,
+          "mask_prompt": prompt,
+          "adjustment_factor": adjustment_factor,
+          "negative_mask_prompt":negative_prompt
+      }
+  )
+  out_items={}
+  for item in output:
+    # https://replicate.com/schananas/grounded_sam/api#output-schema
+    print(item)
+    out_items[os.path.basename(item).split(".")[0]]=item
+  return out_items 
+
+
+def replicate_zest(img,material_img="https://replicate.delivery/pbxt/Kl23gJODaW7EuxrDzBG9dcgqRdMaYSWmBQ9UexnwPiL7AnIr/3.jpg"):
+  if type(img)!=str:
+    img=image_to_base64(img)
+  if type(material_img)!=str:
+    material_img=image_to_base64(material_img)
+
+  output = replicate.run(
+      "camenduru/zest:11abc0a411459327938957581151c642dd1bee4cefe443a9a63b230c4fbc0952",
+      input={
+          "input_image": img,
+          "material_image":material_img
+      }
+  ) 
+  print(output)
+  return output
+
+
+from src.utils import resize_image,find_closest_valid_dimension
+
+
+light_source_options=[
+    "Use Background Image",
+    "Left Light",
+    "Right Light",
+    "Top Light",
+    "Bottom Light",
+    "Ambient"
+]
+
+
+def replicate_iclight_BG(img,prompt,bg_img,light_source="Use Background Image",
+        negative_prompt="(deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime, mutated hands and fingers:1.4), (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, disconnected limbs, mutation, mutated, ugly, disgusting, amputation"
+  ):
+
+    assert light_source in light_source_options, "Please select a correct ligh source option"
+   #if type(img)!=str:
+   # img=open_image_from_url(img)
+   #if type(bg_img)!=str:
+   # bg_img=open_image_from_url(bg_img)
+    
+    width, height = img.size
+    print(width,height)
+    #print()
+    target_width = find_closest_valid_dimension(width)
+    target_height = find_closest_valid_dimension(height)
+    resized_img = resize_image(img, target_width, target_height)
+    img=image_to_base64(resized_img)
+
+    #if light_source=="Use Background Image":
+    bg_width, bg_height = bg_img.size
+    target_width = find_closest_valid_dimension(bg_width)
+    target_height = find_closest_valid_dimension(bg_height)
+    print(bg_img)
+    bg_img = resize_image(bg_img, target_width, target_height)
+    bg_img=image_to_base64(bg_img)
+
+  
+    output=replicate.run(
+       "zsxkib/ic-light-background:60015df78a8a795470da6494822982140d57b150b9ef14354e79302ff89f69e3",
+    input={
+        "cfg": 2,
+        "steps": 25,
+        "width": width,
+        "height": height,
+        "prompt": prompt,
+        "light_source": light_source,
+        "highres_scale": 1.5,
+        "output_format": "png",
+        "subject_image": img,
+        "compute_normal": False,
+        "output_quality": 100,
+        "appended_prompt": "best quality",
+        "highres_denoise": 0.5,
+        "negative_prompt": "lowres, bad anatomy, bad hands, cropped, worst quality",
+        "background_image": bg_img,
+        "number_of_images": 1
+    }
+
+    )
+    return  output[0]

src/utils.py

@@ -12,6 +12,53 @@ import json
 import gradio as gr
 
 
+def resize_image(img, target_width, target_height):
+  """Resizes an image while maintaining aspect ratio.
+
+  Args:
+    img: The PIL Image object to resize.
+    target_width: The desired width.
+    target_height: The desired height.
+
+  Returns:
+    The resized PIL Image object.
+  """
+  width, height = img.size
+  aspect_ratio = width / height
+
+  # Calculate new dimensions based on aspect ratio and target dimensions
+  if width > height:
+    new_width = target_width
+    new_height = int(new_width / aspect_ratio)
+  else:
+    new_height = target_height
+    new_width = int(new_height * aspect_ratio)
+
+  # Resize the image
+  resized_img = img.resize((new_width, new_height))
+  return resized_img
+
+# Example usage:
+# Assuming img is your PIL Image object
+# target_width = 512
+# target_height = 512
+# resized_img = resize_image(img, target_width, target_height)
+# resized_img.show()
+
+def find_closest_valid_dimension(dimension, valid_dimensions= [256, 320, 384, 448, 512, 576, 640, 704, 768, 832, 896, 960, 1024]):
+  """Finds the closest valid dimension from a list of valid dimensions.
+
+  Args:
+    dimension: The target dimension.
+    valid_dimensions: A list of valid dimensions.
+
+  Returns:
+    The closest valid dimension.
+  """
+  closest_dimension = min(valid_dimensions, key=lambda x: abs(x - dimension))
+  return closest_dimension
+
+
 def convert_to_pil(img):
     if isinstance(img, np.ndarray):
         img = Image.fromarray(img)