Remove hardcoded variables in mesh_sam + fix small README typos

P3-SAM/README.md

@@ -17,9 +17,9 @@ Segmenting 3D assets into their constituent parts is crucial for enhancing 3D un
 ![Teaser](./images/teaser.jpg)
 
 ### TODO List 
-- [X] Realse the paper.
-- [X] Realse the code.
-- [X] Realse the pre-trained models.
+- [X] Release the paper.
+- [X] Release the code.
+- [X] Release the pre-trained models.
 
 ### Install 
 1.  We recommend using a virtual environment to install the required packages. 
@@ -38,7 +38,7 @@ Segmenting 3D assets into their constituent parts is crucial for enhancing 3D un
 
 ### Inference
 1. Our demo will automatically download the pre-trained models from huggingface. You can also download `p3sam.ckpt` manually from the [link](https://huggingface.co/tencent/Hunyuan3D-Part) and put it in the `weights` folder.
-2. Run the following command to automantically generate the masks:
+2. Run the following command to automatically generate the masks:
     ```
     cd demo
     python auto_mask.py --ckpt_path ../weights/last.ckpt --mesh_path assets/1.glb --output_path results/1
@@ -53,7 +53,7 @@ Segmenting 3D assets into their constituent parts is crucial for enhancing 3D un
     python app.py --ckpt_path ../weights/last.ckpt --data_dir assets
     ```
     ![APP](./images/app.gif)
-4. Or you can run the following command to open a gradio app to automantically generate the masks:
+4. Or you can run the following command to open a gradio app to automatically generate the masks:
     ```
     cd demo
     python gradio_demo.py 

P3-SAM/demo/auto_mask.py

@@ -776,8 +776,6 @@ def mesh_sam(
     if show_info:
         print(f"点数：{mesh.vertices.shape[0]} 面片数：{mesh.faces.shape[0]}")
 
-    point_num = 100000
-    prompt_num = 400
     with Timer("获取邻接面片"):
         face_adjacency = mesh.face_adjacency
     with Timer("处理邻接面片"):

P3-SAM/demo/auto_mask_no_postprocess.py

@@ -445,8 +445,6 @@ def mesh_sam(
     if show_info:
         print(f"点数：{mesh.vertices.shape[0]} 面片数：{mesh.faces.shape[0]}")
 
-    point_num = 100000
-    prompt_num = 400
 
     with Timer("采样点云"):
         _points, face_idx = trimesh.sample.sample_surface(mesh, point_num, seed=seed)

XPart/README.md

@@ -15,9 +15,9 @@ Generating 3D shapes at part level is pivotal for downstream applications such a
 ![Teaser](./assets/teaser.jpg)
 
 ### TODO List 
-- [X] Realse the paper.
-- [X] Realse the code.
-- [ ] Realse the pre-trained models.
+- [X] Release the paper.
+- [X] Release the code.
+- [ ] Release the pre-trained models.
 
 ### Install 
 1.  We recommend using a virtual environment to install the required packages. 
@@ -33,7 +33,7 @@ Generating 3D shapes at part level is pivotal for downstream applications such a
 
 ### Inference
 1. Download the pre-trained models (TODO)
-2. Run the following command to automantically generate the masks:
+2. Run the following command to automatically generate the masks:
     ```
     python demo.py --config partgen/config/infer.yaml --mesh_path ./data/test.glb --output_path ./results/
     ```

XPart/partgen/bbox_estimator/auto_mask_api.py

@@ -823,8 +823,6 @@ def mesh_sam(
     if show_info:
         print(f"点数：{mesh.vertices.shape[0]} 面片数：{mesh.faces.shape[0]}")
 
-    point_num = 100000
-    prompt_num = 400
     with Timer("获取邻接面片"):
         # 获取邻接面片
         face_adjacency = mesh.face_adjacency