open-mesh-recon 0.0.1.0

The OpenMesh package version 0.0.1 a first version of this package have the ability to work with 3D resconstruction models. Its not completed yet. wait for it.

Welcome to the Open Mesh package pypi page

This package includes two classes for this version (0.0.1) and it's so easy to use them.

What does Open Mesh do

This package generated for those guys who are trying to use 3D human reconstruction models so easy.

How to use Open Mesh

Here you can see a step-by-step tutorial of how you can use the package to work with 3D reconstruction models (for humans) so easy!

Step 1: Project structure

In the following content you can see the structure of the project how to use Open Mesh

Its recommended to use a virtual environment.

Setting up environment

Just run the following commands to install virtualenv and generate a new environment for Open Mesh

pip install virtualenv

virtualenv venv

You will have a new folder called venv in your directory you were with your terminal

venv\Scripts\activate # For windows users

source venv/bin/activate # For linux/macos users

After this you should see a (venv) at the first of your terminal path.

Step 2: Installation

After the virtual environment generated and activated by the terminal you just need to run:

pip install open-mesh-recon

and then it will start downloading all the requirements and the whole package and then install it.

your_project_name/
|
├─── venv/
|
├─── your_main_file.py
|
├─── input_images/
|    |
|    └─── human_image_1.png
|    |
|    └─── human_image_2.png
|    |
|    └─── human_image_3.png
|    |
|    └─── ...

Step 3: Main python script

In the following content you should write your main.py script by the tutorial.

from omesh import MeshFactory, PyLog

MeshFactory

MeshFactory is the core class which includes all the functions to control the whole process.

In the following script you can see all the functions should run in order.

mesh_factory = MeshFactory(mesh_id_min=10, mesh_id_max=20)

mesh_factory.remove_background()

mesh_factory.extract_keypoints()

mesh_factory.run_model()

mesh_factory.save_mesh()

mesh_factory.render_spinning_video()

The object of this class has a mesh_id which generated randomly between the range passed as mesh_id_min and mesh_id_max

The program will automaticly detect the device should be cuda or cpu using torch.cuda.is_available() method and the pass it with the run_model() function.

The input directory can be changed by editing input_dir argument.

# Example:
mf = MeshFactory(input_dir='your_input_directory')
# It is setted to 'input_images' by default which defined in folder structure

The last argument of this class is output_dir which determines the output directory of the process which setted to results by default. You do not need to generate the output directory.

PyLog

PyLog is a simple logging class for better debugging. It includes 4 functions info, warning, error, succeed. The functions info, warning and succeed are the same just with different headers. You just need to pass a message to all of these functions. The error function stops the program directly after the message displayed.

You can see a simple worker for PyLog below + the output

log = PyLog()
log.info("This is a simple info")
log.warning("The addon package not installed!")
log.error("Failed to run model!")
log.succeed("Finished Processing")

The output might looks like this:

 --- INFO ---: This is a simple info
 --- WARNING ---: The addon package not installed!
 --- ERROR ---: Failed to run model!

Tip: Actually the last successful message will not appear beacause the previous error message will stop the program so the rest of the program will not execute.

log.succeed("Program Executed!")

The output of a succeed function looks like this:

 --- SUCCESSFUL ---: Program Executed!

Step 4: Output

Once you run your main.py file you will be able to see the results folder (or the name you specified as 'output_dir'). For each image exists on the input directory there is a removed background image and a 2d keypoints including body landmarks with openpose json format.

You might see the mesh_mesh id.obj file + the .mp4 spinning video rendered once the process finished. The program logs for every step and you would see errors if something was not expected.

Notice: This release (0.0.1) is not completed and does not include the implemented functions run_model, save_mesh and render_spinning_video, a pre-release version.

Future Versions

In the future releases I might implement the functions not defined. That functions just log their process but do not work.