feat: Enhance LossManager with advanced loss computation methods

- Added `position_loss` method to calculate manifold loss using mean absolute value
- Implemented `normals_loss` method to compute normal vector loss with gradient calculation
- Updated `train.py` to log input and output tensor shapes during training
- Modified network architecture in `train.py` by increasing hidden layer dimensions from [64, 64, 64] to [256, 256, 256]

code/conversion/loss.py

@@ -1,11 +1,43 @@
-
-
+import torch
 
 
 class LossManager:
     def __init__(self):
         pass
 
+    def position_loss(self, outputs):
+        """
+        计算流型损失的逻辑
+
+        :param outputs: 模型的输出
+        :return: 计算得到的流型损失值     
+        """
+
+        # 计算流型损失（这里使用均方误差作为示例）
+        manifold_loss = (outputs.abs()).mean()  # 计算流型损失
+        return manifold_loss
+    
+    def normals_loss(self, cur_data: torch.Tensor, mnfld_pnts: torch.Tensor, all_fi: torch.Tensor, patch_sup: bool) -> torch.Tensor:
+        """
+        计算法线损失
+
+        :param cur_data: 当前数据，包含法线信息
+        :param mnfld_pnts: 流型点
+        :param all_fi: 所有流型预测值
+        :param patch_sup: 是否支持补丁
+        :return: 计算得到的法线损失
+        """
+        # 提取法线
+        normals = cur_data[:, -self.d_in:]  
+
+        # 计算分支梯度
+        branch_grad = gradient(mnfld_pnts, all_fi[:, 0])  # 计算分支梯度
+
+        # 计算法线损失
+        normals_loss = (((branch_grad - normals).abs()).norm(2, dim=1)).mean()  # 计算法线损失
+
+        return self.normals_lambda * normals_loss  # 返回加权后的法线损失
+
     def compute_loss(self, outputs):
         """
         计算流型损失的逻辑

code/conversion/train.py

@@ -23,7 +23,7 @@ class NHREPNet_Training:
         
         # 初始化模型
         d_in = 6  # 输入维度，例如 3D 坐标
-        dims_sdf = [64, 64, 64]  # 隐藏层维度
+        dims_sdf = [256, 256, 256]  # 隐藏层维度
         csg_tree, _ = self.dataset.get_csg_tree()
         self.loss_manager = LossManager()
         self.model = NHRepNet(d_in, dims_sdf, csg_tree).to(self.device)  # 实例化模型并移动到设备
@@ -51,9 +51,11 @@ class NHREPNet_Training:
 
         # 获取输入数据
         input_data = self.dataset.get_data().to(self.device)  # 获取数据并移动到设备
+        logger.info(f"输入数据: {input_data.shape}")
         
         # 前向传播
         outputs = self.model(input_data)  # 使用模型进行前向传播
+        logger.info(f"输出数据: {outputs.shape}")
         
         # 计算损失
         loss = self.loss_manager.compute_loss(outputs)  # 计算损失
@@ -68,17 +70,7 @@ class NHREPNet_Training:
         logger.info(f'Epoch [{epoch}/{self.nepochs}], Average Loss: {avg_loss:.4f}')
         self.writer.add_scalar('Loss/train', avg_loss, epoch)  # 记录损失到 TensorBoard
 
-    def compute_loss(self, outputs):
-        """
-        计算流型损失的逻辑
 
-        :param outputs: 模型的输出
-        :return: 计算得到的流型损失值     
-        """
-
-        # 计算流型损失（这里使用均方误差作为示例）
-        manifold_loss = (outputs.abs()).mean()  # 计算流型损失
-        return manifold_loss
 
 if __name__ == "__main__":
     data_dir = '../data/input_data'  # 数据目录