Skinned Mesh Processor

A C++ application for loading, skinning, and saving 3D meshes using bone weights and pose data. This tool processes .obj files with skinning information provided via JSON files, applies transformations using skeletal animation techniques, and outputs the result as a new .obj file. Built with performance in mind, it leverages OpenMP for parallel processing and GLM for matrix operations.

Features

• Loads 3D meshes from .obj files using tinyobjloader.
• Applies skinning transformations based on bone weights, inverse bind poses, and new poses from JSON files.
• Supports up to 4 bone influences per vertex.
• Parallelized skinning with OpenMP for improved performance.
• Saves the skinned mesh to a new .obj file.
• Includes error handling for file I/O and data consistency.

Dependencies

• tinyobjloader: Loads .obj files (v2.0.0rc13).
• nlohmann/json: Parses JSON data (v3.11.3).
• glm: Handles vector and matrix math (v1.0.1).
• OpenMP: Enables parallel processing (requires compiler support, e.g., -fopenmp for GCC/Clang).

Prerequisites

• CMake 3.15 or higher
• A C++17-compatible compiler (e.g., GCC, Clang, MSVC)
• Git (for cloning and fetching dependencies)

Building the Project

1. Clone the Repository

   git clone https://github.com/<your-username>/skinned-mesh-processor.git
   cd skinned-mesh-processor

2. Create a Build Directory

mkdir build
cd build

3. Run CMake

cmake ..

4. Build the Project

   • Linux/macOS:

   make

   • Windows (Visual Studio): Open the generated ``.sln` file in Visual Studio and build, or use:

   cmake --build . --config Release

   The executable will be placed in `build/bin/``.

Usage

Run the program with a folder containing the required input files:

./bin/ObjSkinning <folder_path>

Input Files

Place these files in the specified <folder_path>:

• test_mesh.obj: The input 3D mesh.
• bone_weight.json: Bone indices and weights for each vertex.
• inverse_bind_pose.json: Inverse bind pose matrices for each bone.
• new_pose.json: New pose matrices for each bone.

Output

The program generates <base_filename>_skinned.obj (e.g., test_mesh_skinned.obj) in the same folder and prints the execution time in microseconds.

Example

./bin/ObjSkinning ./test

• Input: data/test_mesh.obj, data/bone_weight.json, etc.
• Output: data/test_mesh_skinned.obj

Performance Notes

• Multithreading: Initial attempts with std::thread were replaced with OpenMP, reducing overhead and improving execution time by a few milliseconds.
• SIMD Potential: Intel SSE intrinsics (immintrin.h) showed promise but were excluded for portability. Future work could explore cross-platform SIMD.
• Granularity: The skinning operation benefits from OpenMP’s lightweight parallelism but may not scale well with coarse thread pools.

Project Structure

skinned-mesh-processor/
├── src/
│   └── main.cpp          # Core implementation
├── CMakeLists.txt        # Build configuration
└── README.md             # This file

Example Output

Below is a comparison of the original and processed meshes:

Original Mesh	Processed Mesh