add controlImage

app.py

@@ -3,23 +3,27 @@ import numpy as np
 
 import spaces
 import torch
-import spaces
 import random
+from image_gen_aux import DepthPreprocessor
 
 from diffusers import FluxFillPipeline
 from PIL import Image
 
+def remove_background(image):
+       # Placeholder function for background removal
+       # Use a library or model like Inspyrenet for actual implementation
+       mask = generate_mask(image)
+       subject = apply_mask(image, mask)
+       return subject, mask
 
 MAX_SEED = np.iinfo(np.int32).max
 MAX_IMAGE_SIZE = 2048
 
 pipe = FluxFillPipeline.from_pretrained("black-forest-labs/FLUX.1-Fill-dev", torch_dtype=torch.bfloat16).to("cuda")
 pipe.load_lora_weights("alvdansen/flux-koda")
-# pipe.enable_sequential_cpu_offload()
-# pipe.enable_fp16()
 pipe.enable_lora()
-# pipe.vae.enable_slicing()
-# pipe.vae.enable_tiling()
+
+processor = DepthPreprocessor.from_pretrained("LiheYoung/depth-anything-large-hf")
 
 def calculate_optimal_dimensions(image: Image.Image):
     # Extract the original dimensions
@@ -67,7 +71,10 @@ def infer(edit_images, prompt, seed=42, randomize_seed=False, width=1024, height
     mask = edit_images["layers"][0]
     if randomize_seed:
         seed = random.randint(0, MAX_SEED)
+
+    controlImage = processor(image)[0].convert("RGB")
     image = pipe(
+        control_image=controlImage,
         prompt=prompt,
         image=image,
         mask_image=mask,
@@ -183,4 +190,4 @@ with gr.Blocks(css=css) as demo:
         outputs = [result, seed]
     )
 
-demo.launch()
+demo.launch()

requirements.txt

@@ -1,3 +1,4 @@
+git+https://github.com/asomoza/image_gen_aux.git
 git+https://github.com/huggingface/diffusers.git
 transformers
 accelerate