update

app.py

@@ -83,7 +83,8 @@ def arbitrary_to_apose(image, seed):
 def apose_to_multiview(apose_img, seed):
     # convert image to PIL.Image
     apose_img = Image.fromarray(apose_img)
-    results, _ = infer_api.genStage2(apose_img, seed, num_levels=1)
+    results = infer_api.genStage2(apose_img, seed, num_levels=1)
+    infer_api.add_results(results)
     return results[0]["images"]
 
 def multiview_to_mesh(images):
@@ -92,9 +93,10 @@ def multiview_to_mesh(images):
 
 def refine_mesh(apose_img, mesh1, mesh2, mesh3, seed):
     apose_img = Image.fromarray(apose_img)
-    _, all_results = infer_api.genStage2(apose_img, seed, num_levels=2)
-    print(all_results.keys())
-    refined = infer_api.genStage4([mesh1, mesh2, mesh3], all_results)
+    results = infer_api.genStage2(apose_img, seed, num_levels=2)
+    infer_api.add_results(results)
+    print(infer_api.results.keys())
+    refined = infer_api.genStage4([mesh1, mesh2, mesh3], infer_api.results)
     return refined
 
 with gr.Blocks(title="StdGEN: Semantically Decomposed 3D Character Generation from Single Images") as demo:

infer_api.py

@@ -381,6 +381,7 @@ class InferAPI:
         self.multiview_configs = multiview_configs
         self.slrm_configs = slrm_configs
         self.refine_configs = refine_configs
+        self.results = {}
 
         # self.canonical_infer = InferCanonicalAPI(self.canonical_configs)
         # self.multiview_infer = InferMultiviewAPI(self.multiview_configs)
@@ -398,6 +399,10 @@ class InferAPI:
 
     def genStage4(self, meshes, imgs):
         return infer_refine(meshes, imgs)
+    
+    def add_results(self, results):
+        for k in results:
+            self.results[k] = results[k]
 
 
 ############## Refine ##############
@@ -810,8 +815,6 @@ def process_im(im):
     im = infer_multiview_image_transforms(im)
     return im
 
-all_results = {}
-
 @spaces.GPU(duration=180)
 def infer_multiview_gen(img, seed, num_levels):
     set_seed(seed)
@@ -824,9 +827,7 @@ def infer_multiview_gen(img, seed, num_levels):
     data["color_prompt_embeddings"] = infer_multiview_color_text_embeds[None, ...]
 
     results = run_multiview_infer(data, infer_multiview_pipeline, infer_multiview_cfg, num_levels=num_levels)
-    for k in results:
-        all_results[k] = results[k]
-    return results, all_results
+    return results
 
 infer_canonicalize_config = {
     'config_path': './configs/canonicalization-infer.yaml',