[Add]add paddle version code of IJACA 2024

Author: lijialin03

jointContribution/IJACA_2024/README.md

@@ -0,0 +1,24 @@
+# IJACA_Code
+The paddle version of the top three in each track of the IJACA 2024 competition.
+Inference codes only now.
+
+## Dataset(To be added)
+Please refer to the .ipynb files in each directory to download the data and set the corresponding parameters.
+
+## Checkpoint(To be added)
+
+## Inference commands
+### aminos
+python infer.py --dataset_dir /your_path/Datasets
+
+### tenfeng
+python infer.py --epochs 69 --milestones 40 50 60 65 68 --gpu_id 0  --depth 5 --hidden_dim 256 --num_slices 32 --batch_size 4 --loss_type 'rl2' --submit --log_dir your_path --training_data_dir /your_path/Dataset/train_track_B_e --testing_data_dir /your_path/Dataset/Testset_track_B_e
+
+### leejt
+python infer.py
+
+### bju
+python infer.py
+
+### zhongzaicanyu
+python infer.py

jointContribution/IJACA_2024/aminos/Extract_mesh/__init__.py

@@ -0,0 +1,121 @@
+import argparse
+import os
+
+import h5py
+import numpy as np
+import paddle
+
+
+def load_ds_trackA_info(file_path, key_list):
+    path_trackA_ds = file_path
+    key_list = np.sort([int(key) for key in key_list])
+    key_list = [str(key) for key in key_list]
+    bounds = np.loadtxt(path_trackA_ds + "/watertight_global_bounds.txt")
+    pressure_mean_std = paddle.to_tensor(
+        data=np.loadtxt(path_trackA_ds + "/train_pressure_min_std.txt")
+    ).to("float32")
+    voxel_mean_std = paddle.to_tensor(
+        data=np.loadtxt(path_trackA_ds + "/voxel_mean_std.txt")
+    ).to("float32")
+    pos_mean_std = np.loadtxt(path_trackA_ds + "/pos_mean_std.txt")
+    normal_mean_std = np.loadtxt(path_trackA_ds + "/normal_mean_std.txt")
+    PN_mean_std = paddle.to_tensor(
+        data=np.concatenate([pos_mean_std, normal_mean_std], axis=-1)
+    ).to("float32")
+    physics_info = {
+        "key_list": key_list,
+        "bounds": bounds,
+        "voxel_mean_std": voxel_mean_std,
+        "pressure_mean_std": pressure_mean_std,
+        "PN_mean_std": PN_mean_std,
+    }
+    return physics_info
+
+
+def load_ds_trackB_info(file_path, key_list):
+    path_trackB_ds = file_path
+    key_list = np.sort([int(key) for key in key_list])
+    key_list = [str(key) for key in key_list]
+    pressure_mean_std = paddle.to_tensor(
+        data=np.loadtxt(path_trackB_ds + "/train_pressure_mean_std.txt")
+    ).to("float32")
+    bounds = np.loadtxt(path_trackB_ds + "/global_bounds.txt")
+    voxel_mean_std = paddle.to_tensor(
+        data=np.loadtxt(path_trackB_ds + "/voxel_mean_std.txt")
+    ).to("float32")
+    PNA_mean_std = paddle.to_tensor(
+        data=np.loadtxt(path_trackB_ds + "/PosNormalArea_mean_std.txt")
+    ).to("float32")
+    PN_mean_std = PNA_mean_std[:, :6]
+    physics_info = {
+        "key_list": key_list,
+        "bounds": bounds,
+        "voxel_mean_std": voxel_mean_std,
+        "pressure_mean_std": pressure_mean_std,
+        "PN_mean_std": PN_mean_std,
+    }
+    return physics_info
+
+
+def load_extra_info(file_path, key_list, track_type="A"):
+    if track_type == "A":
+        physics_info = load_ds_trackA_info(file_path, key_list)
+    else:
+        physics_info = load_ds_trackB_info(file_path, key_list)
+    return physics_info
+
+
+def add_physics_info_to_group(group, physics_info):
+    for key, value in physics_info.items():
+        group.create_dataset(key, data=value)
+
+
+def merge_h5_files(fileA_path, fileB_path, merged_file_path):
+    with h5py.File(fileA_path, "r") as fileA, h5py.File(
+        fileB_path, "r"
+    ) as fileB, h5py.File(merged_file_path, "w") as merged_file:
+        key_list_A = list(fileA.keys())
+        key_list_B = list(fileB.keys())
+        physics_info_A = load_extra_info(
+            os.path.dirname(fileA_path), key_list_A, track_type="A"
+        )
+        physics_info_B = load_extra_info(
+            os.path.dirname(fileB_path), key_list_B, track_type="B"
+        )
+        for key in fileA.keys():
+            group = fileA[key]
+            new_key = "A_" + key
+            merged_file.copy(group, new_key)
+            add_physics_info_to_group(merged_file[new_key], physics_info_A)
+        for key in fileB.keys():
+            group = fileB[key]
+            new_key = "B_" + key
+            merged_file.copy(group, new_key)
+            add_physics_info_to_group(merged_file[new_key], physics_info_B)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(
+        description="train / test a pytorch model to predict frames"
+    )
+    parser.add_argument(
+        "--A_dir",
+        default="/home/xiaoli/project/3D-ShapeNet-car/src/Dataset/converted_dataset/trackA/test.h5",
+        type=str,
+        help="",
+    )
+    parser.add_argument(
+        "--B_dir",
+        default="/home/xiaoli/project/3D-ShapeNet-car/src/Dataset/converted_dataset/trackB/test.h5",
+        type=str,
+        help="",
+    )
+    parser.add_argument(
+        "--C_dir",
+        default="/home/xiaoli/project/3D-ShapeNet-car/src/Dataset/converted_dataset/trackC/k1.h5",
+        type=str,
+        help="",
+    )
+    params = parser.parse_args()
+    merge_h5_files(params.A_dir, params.B_dir, params.C_dir)
+print("done")