Skip to content

Instantly share code, notes, and snippets.

@JaySmithWpg
Created December 9, 2022 21:37
Show Gist options
  • Save JaySmithWpg/0dfb716fef567b5fbe8fbebf38dd1101 to your computer and use it in GitHub Desktop.
Save JaySmithWpg/0dfb716fef567b5fbe8fbebf38dd1101 to your computer and use it in GitHub Desktop.
Hack to load custom depth map
diff --git a/modules/processing.py b/modules/processing.py
index 0417ffc..c2bde57 100644
--- a/modules/processing.py
+++ b/modules/processing.py
@@ -154,15 +154,17 @@ class StableDiffusionProcessing():
return image_conditioning
def depth2img_image_conditioning(self, source_image):
- # Use the AddMiDaS helper to Format our source image to suit the MiDaS model
- transformer = AddMiDaS(model_type="dpt_hybrid")
- transformed = transformer({"jpg": rearrange(source_image[0], "c h w -> h w c")})
- midas_in = torch.from_numpy(transformed["midas_in"][None, ...]).to(device=shared.device)
- midas_in = repeat(midas_in, "1 ... -> n ...", n=self.batch_size)
+ depth_img = Image.open("/home/jay/Pictures/AI/Turrent Room/depth.png")
+ depth_img = depth_img.convert("L")
+ depth_img = np.expand_dims(depth_img, axis=0)
+ depth_img = np.expand_dims(depth_img, axis=0).repeat(self.batch_size, axis=0)
+ depth_img = torch.from_numpy(depth_img)
+ depth_img = 2. * depth_img - 1.
+ depth_img = depth_img.to(shared.device)
conditioning_image = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(source_image))
conditioning = torch.nn.functional.interpolate(
- self.sd_model.depth_model(midas_in),
+ depth_img,
size=conditioning_image.shape[2:],
mode="bicubic",
align_corners=False,
@loboere
Copy link

loboere commented Dec 13, 2022

it would be great to have an interface as "drag and drop custom depth map" in the interface of automatic 11111

@xrstack
Copy link

xrstack commented Feb 1, 2023

+1, would be amazing to have this in the the a1111 interface to select a synthetic depth map, instead of computing one using MiDaS

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment