update

app.py

@@ -47,7 +47,7 @@ This is official demo for our CVPR 2025 paper <a href="">StdGEN: Semantic-Decomp
 
 Code: <a href='https://github.com/hyz317/StdGEN' target='_blank'>GitHub</a>. Paper: <a href='https://arxiv.org/abs/2411.05738' target='_blank'>ArXiv</a>.
 
-❗️❗️❗️**Important Notes:** This is only a **PREVIEW** version with lower quality. We perform only color back-projection to clothes and hair. Please refer to GitHub repo for complete version.
+❗️❗️❗️**Important Notes:** This is only a **PREVIEW** version with **coarse precision geometry and texture** due to gradio use. We skip some refinement process and perform only color back-projection to clothes and hair. Please refer to GitHub repo for complete version.
 1. Refinement stage takes about ~2.5min, and the mesh result may possibly delayed due to the server load, please wait patiently.
 
 2. You can upload any reference image (with or without background), A-pose images are also supported (white bkg required). If the image has an alpha channel (transparency), background segmentation will be automatically performed. Alternatively, you can pre-segment the background using other tools and upload the result directly.  
@@ -112,7 +112,7 @@ with gr.Blocks(title="StdGEN: Semantically Decomposed 3D Character Generation fr
             )      
             seed_input = gr.Number(
                 label="Seed",
-                value=42,
+                value=50,
                 precision=0,
                 interactive=True
             )
@@ -127,7 +127,7 @@ with gr.Blocks(title="StdGEN: Semantically Decomposed 3D Character Generation fr
             )
             seed_input2 = gr.Number(
                 label="Seed",
-                value=42,
+                value=50,
                 precision=0,
                 interactive=True
             )

infer_multiview.py

@@ -227,7 +227,6 @@ def load_multiview_pipeline(cfg):
         cfg.pretrained_path,
         torch_dtype=torch.float16,)
     pipeline.unet.enable_xformers_memory_efficient_attention()
-    import pdb; pdb.set_trace()
     if torch.cuda.is_available():
         pipeline.to(device)
     return pipeline

refine/mesh_refine.py

@@ -313,7 +313,7 @@ def geo_refine_2(vertices, faces, fixed_v=None):
     meshes = simple_clean_mesh(to_pyml_mesh(vertices, faces), apply_smooth=True, stepsmoothnum=2, apply_sub_divide=False, sub_divide_threshold=0.25)
     simp_vertices, simp_faces = meshes.verts_packed(), meshes.faces_packed()
     vertices, faces = simp_vertices.detach().cpu().numpy(), simp_faces.detach().cpu().numpy()
-    vertices, faces = trimesh.remesh.subdivide(vertices, faces)
+    # vertices, faces = trimesh.remesh.subdivide(vertices, faces)
     if fixed_v is not None:
         vertices, faces = trimesh.remesh.subdivide(vertices, faces)
     return vertices, faces