Refactor mesh generation in app.py and update remeshing parameters in inference.py

- Removed trimesh and KDTree functionality from app.py for simplified mesh handling.
- Adjusted remesh size and iterations in the generate3d function to enhance mesh quality.

app.py

@@ -18,8 +18,6 @@ import json
 import argparse
 import requests
 import tempfile
-import trimesh
-from scipy.spatial import cKDTree
 
 from model import CRM
 from inference import generate3d
@@ -278,19 +276,4 @@ with gr.Blocks() as demo:
         inputs=inputs,
         outputs=outputs,
     )
-    demo.queue().launch()
-
-    # After mesh generation, fill holes
-    mesh = trimesh.Trimesh(vertices=verts, faces=faces)
-    mesh.fill_holes()
-
-    # Find new vertices (those not in the original set)
-    if mesh.vertices.shape[0] > old_vertices.shape[0]:
-        new_vtx = mesh.vertices[old_vertices.shape[0]:]
-        # Use KDTree to find nearest old vertex for each new vertex
-        tree = cKDTree(old_vertices)
-        dists, idxs = tree.query(new_vtx)
-        new_colors = old_colors[idxs]
-        # Concatenate old and new colors
-        all_colors = np.vstack([old_colors, new_colors])
-        mesh.visual.vertex_colors = all_colors
+    demo.queue().launch()

inference.py

@@ -63,7 +63,7 @@ def generate3d(model, rgb, ccm, device):
     verts, faces = clean_mesh(
         data_config['verts'].squeeze().cpu().numpy().astype(np.float32),
         data_config['faces'].squeeze().cpu().numpy().astype(np.int32),
-        repair=False, remesh=True, remesh_size=0.005, remesh_iters=1
+        repair=False, remesh=True, remesh_size=0.01, remesh_iters=2
     )
     data_config['verts'] = torch.from_numpy(verts).to(device).contiguous()
     data_config['faces'] = torch.from_numpy(faces).to(device).contiguous()