Add ComfyUI Progress Bar for Zero123++, add MVDream output camposes

README.md

@@ -1,5 +1,7 @@
 # ComfyUI-3D-Pack
- An extensive node suite that enables ComfyUI to process 3D inputs (Mesh & UV Texture, etc) using cutting edge algorithms (3DGS, NeRF, Differentiable Rendering, SDS/VSD Optimization, etc.)
+**Make ComfyUI generates 3D assets as good & convenient as it generates image/video!**
+<br>
+This is an extensive node suite that enables ComfyUI to process 3D inputs (Mesh & UV Texture, etc.) using cutting edge algorithms (3DGS, NeRF, etc.) and models (InstantMesh, CRM, TripoSR, etc.)
 
 <span style="font-size:1.5em;">
 <a href=#currently-support>Features</a> &mdash;
@@ -10,8 +12,6 @@
 <a href=#supporters>Supporters</a>
 </span>
 
-### Note: this project is still a WIP
-
 ## Currently support:
 - For use case please check [Example Workflows](./_Example_Workflows/). [**Last update: 23/05/2024**]
   - **Note:** you need to put [Example Inputs Files & Folders](_Example_Workflows/_Example_Inputs_Files/) under ComfyUI Root Directory\ComfyUI\input folder before you can run the example workflow
@@ -180,7 +180,7 @@ install_windows_portable_win_py311_cu121.bat
 ```
 
 ### Install Method 1: Using Miniconda(Works on Windows & Linux & Mac)
-***Note: [In some edge cases Miniconda fails Anaconda could fix the issue](https://github.com/MrForExample/ComfyUI-3D-Pack/issues/49)***
+***Note: [In some edge cases Miniconda fails but Anaconda could fix the issue](https://github.com/MrForExample/ComfyUI-3D-Pack/issues/49)***
 
 #### Setup with Miniconda:
 First download [Miniconda](https://docs.conda.io/projects/miniconda/en/latest/) (*One of the best way to manage a clean and separated python envirments*)
@@ -230,6 +230,7 @@ pip install -r requirements_post.txt
 
 **Plus:**<br>
 - For those who want to run it inside Google Colab, you can check the [install instruction from @lovisdotio](https://github.com/MrForExample/ComfyUI-3D-Pack/issues/13)
+- You can find some of the pre-build wheels for Linux here: [remsky/ComfyUI3D-Assorted-Wheels](https://github.com/remsky/ComfyUI3D-Assorted-Wheels)
 
 #### Install and run with docker:
 

nodes.py

@@ -1487,9 +1487,11 @@ class MVDream_Model:
     
     RETURN_TYPES = (
         "IMAGE",
+        "ORBIT_CAMPOSES",   # [orbit radius, elevation, azimuth, orbit center X, orbit center Y, orbit center Z]
     )
     RETURN_NAMES = (
         "multiview_images",
+        "orbit_camposes",
     )
     FUNCTION = "run_mvdream"
     CATEGORY = "Comfy3D/Algorithm"
@@ -1521,7 +1523,14 @@ class MVDream_Model:
         mv_images = mvdream_pipe(prompt, reference_image, generator=generator, negative_prompt=prompt_neg, guidance_scale=mv_guidance_scale, num_inference_steps=num_inference_steps, elevation=elevation)
         mv_images = torch.from_numpy(np.stack([mv_images[1], mv_images[2], mv_images[3], mv_images[0]], axis=0)).float() # [4, H, W, 3], float32
         
-        return (mv_images, )
+        azimuths = [0, 90, 180, -90]
+        elevations = [0, 0, 0, 0]
+        radius = [4.0] * 4
+        center = [0.0] * 4
+
+        orbit_camposes = [azimuths, elevations, radius, center, center, center]
+
+        return (mv_images, orbit_camposes)
     
 class Load_Large_Multiview_Gaussian_Model:
     
@@ -2246,8 +2255,6 @@ class InstantMesh_Reconstruction_Model:
     @torch.no_grad()
     def run_LRM(self, lrm_model, multiview_images, orbit_camera_poses, orbit_camera_fovy, texture_resolution):
 
-        multiview_images
-
         images = multiview_images.permute(0, 3, 1, 2).unsqueeze(0).to(DEVICE)   # [N, H, W, 3] -> [1, N, 3, H, W]
         images = v2.functional.resize(images, 320, interpolation=3, antialias=True).clamp(0, 1)
 

pyproject.toml

@@ -1,6 +1,9 @@
 [project]
 name = "comfyui-3d-pack"
-description = "An extensive node suite that enables ComfyUI to process 3D inputs (Mesh & UV Texture, etc) using cutting edge algorithms (3DGS, NeRF, etc.)\nNOTE: Pre-built python wheels can be download from [a/https://github.com/remsky/ComfyUI3D-Assorted-Wheels](https://github.com/remsky/ComfyUI3D-Assorted-Wheels)"
+description = "
+    Make ComfyUI generates 3D assets as good & convenient as it generates image/video!
+    This is an extensive node suite that enables ComfyUI to process 3D inputs (Mesh & UV Texture, etc.) using cutting edge algorithms (3DGS, NeRF, etc.) and models (InstantMesh, CRM, TripoSR, etc.)
+"
 version = "1.0.0"
 license = "LICENSE"
 dependencies = ["# base", "cmake", "ninja", "# computational libraries", "numpy", "einops", "scipy", "kornia", "opencv-python", "pillow", "roma", "nerfacc>=0.5.3", "PyMCubes", "scikit-learn", "# for use ML models", "diffusers>=0.26.1", "transformers>=4.36.2", "safetensors", "open_clip_torch", "# for training differentiable tensors", "pytorch_msssim", "# for process images & videos", "imageio", "imageio-ffmpeg", "matplotlib", "# for dmtet and mesh import & export", "trimesh", "plyfile", "pygltflib", "xatlas", "pymeshlab", "# configs & extra", "torchtyping", "tqdm", "jaxtyping", "packaging", "OmegaConf", "pyhocon"]

zero123plus/pipeline.py

@@ -26,6 +26,11 @@ from diffusers.image_processor import VaeImageProcessor
 from diffusers.models.attention_processor import Attention, AttnProcessor, XFormersAttnProcessor, AttnProcessor2_0
 from diffusers.utils.import_utils import is_xformers_available
 
+import comfy.utils
+
+def callback_update_comfy_bar(pipe, step_index, timestep, callback_kwargs):
+        pipe.comfy_pbar.update_absolute(step_index + 1)
+        return callback_kwargs
 
 def to_rgb_image(maybe_rgba: Image.Image):
     if maybe_rgba.mode == 'RGB':
@@ -380,6 +385,9 @@ class Zero123PlusPipeline(diffusers.StableDiffusionPipeline):
         cak = dict(cond_lat=cond_lat)
         if hasattr(self.unet, "controlnet"):
             cak['control_depth'] = depth_image
+
+        self.comfy_pbar = comfy.utils.ProgressBar(num_inference_steps)
+
         latents: torch.Tensor = super().__call__(
             None,
             *args,
@@ -391,6 +399,8 @@ class Zero123PlusPipeline(diffusers.StableDiffusionPipeline):
             output_type='latent',
             width=width,
             height=height,
+            callback_on_step_end=callback_update_comfy_bar,
+            callback_on_step_end_tensor_inputs=[],
             **kwargs
         ).images
         latents = unscale_latents(latents)