delele unnecessary dependency

app.py

@@ -128,6 +128,8 @@ def main(cfg: ProjectConfig):
     # Setup model
     runner = DemoRunner(cfg)
 
+    # runner = None # without model initialization, it shows one line of thumbnail
+
     # Setup interface
     demo = gr.Blocks(title="HDM Interaction Reconstruction Demo")
     with demo:

dataset/__init__.py

@@ -15,16 +15,17 @@ from pytorch3d.implicitron.dataset.json_index_dataset_map_provider_v2 import (
 from pytorch3d.implicitron.tools.config import expand_args_fields
 from pytorch3d.renderer.cameras import CamerasBase
 from torch.utils.data import DataLoader
+from pytorch3d.datasets import R2N2, collate_batched_meshes
 
 from configs.structured import CO3DConfig, DataloaderConfig, ProjectConfig, Optional
-from .exclude_sequence import EXCLUDE_SEQUENCE, LOW_QUALITY_SEQUENCE
 from .utils import DatasetMap
-from .r2n2_my import R2N2Sample, collate_batched_meshes
+
 
 
 def get_dataset(cfg: ProjectConfig):
     
     if cfg.dataset.type == 'co3dv2':
+        from .exclude_sequence import EXCLUDE_SEQUENCE, LOW_QUALITY_SEQUENCE
         dataset_cfg: CO3DConfig = cfg.dataset
         dataloader_cfg: DataloaderConfig = cfg.dataloader
 
@@ -100,6 +101,7 @@ def get_dataset(cfg: ProjectConfig):
         dataloader_val.batch_sampler.drop_last = False
     elif cfg.dataset.type == 'shapenet_r2n2':
         # from ..configs.structured import ShapeNetR2N2Config
+        from .r2n2_my import R2N2Sample
         dataset_cfg: ShapeNetR2N2Config = cfg.dataset
         # for k in dataset_cfg:
         #     print(k)

dataset/utils.py

@@ -0,0 +1,47 @@
+from dataclasses import dataclass
+from typing import Dict, Iterable, Optional
+
+import torch
+import numpy as np
+
+def show_item(item: Dict):
+    for key in item.keys():
+        value = item[key]
+        if torch.is_tensor(value) and value.numel() < 5:
+            value_str = value
+        elif torch.is_tensor(value):
+            value_str = value.shape
+        elif isinstance(value, str):
+            value_str = ('...' + value[-52:]) if len(value) > 50 else value
+        elif isinstance(value, dict):
+            value_str = str({k: type(v) for k, v in value.items()})
+        else:
+            value_str = type(value)
+        print(f"{key:<30} {value_str}")
+
+
+def normalize_to_zero_one(x: torch.Tensor):
+    return (x - x.min()) / (x.max() - x.min())
+    
+
+def default(x, d):
+    return d if x is None else x
+
+
+@dataclass
+class DatasetMap:
+    train: Optional[Iterable] = None
+    val: Optional[Iterable] = None
+    test: Optional[Iterable] = None
+
+
+
+def create_grid_points(bound=1.0, res=128):
+    x_ = np.linspace(-bound, bound, res)
+    y_ = np.linspace(-bound, bound, res)
+    z_ = np.linspace(-bound, bound, res)
+
+    x, y, z = np.meshgrid(x_, y_, z_)
+    # print(x.shape, y.shape) # (res, res, res)
+    pts = np.concatenate([y.reshape(-1, 1), x.reshape(-1, 1), z.reshape(-1, 1)], axis=-1)
+    return pts

requirements.txt

@@ -13,4 +13,5 @@ tqdm
 transformers
 wandb
 trimesh
-pytorch3d
+gradio
+"git+https://github.com/facebookresearch/pytorch3d.git"