Pix3D

Front readme image

The focus of this example is the bproc.loader.load_pix3d(), which can be used to load objects from the Pix3D dataset.

We provide a script to download the .obj files, please see the scripts folder.

Usage

Execute in the BlenderProc main directory:

blenderproc run examples/datasets/pix3d/main.py <PATH_TO_Pix3D> examples/datasets/pix3d/output

  • examples/datasets/pix3d/main.py: path to the python file with pipeline configuration.

  • <PATH_TO_Pix3D>: path to the downloaded pix3d dataset, get it [here](http://pix3d.csail.mit.edu/)

  • examples/datasets/pix3d/output: path to the output directory.

Visualization

In the output folder you will find a series of .hdf5 containers. These can be visualized with the script:

blenderproc vis hdf5 examples/datasets/pix3d/output/*.hdf5

Steps

  • The Pix3DLoader loads all the object paths with the category = bed.

  • One of them is now randomly selected and loaded

Python file (main.py)

Pix3DLoader

# Load Pix3D objects from type table into the scene
objs = bproc.loader.load_pix3d(data_path=args.pix_path, used_category="bed")

This loads a Pix3D Object, it only needs the path to the Pix3D folder, which is saved in args.pix_path. The used_category = bed means a random bed model will be loaded.

The position will be in the center of the scene.

CameraSampler

# Find point of interest, all cam poses should look towards it
poi = bproc.object.compute_poi(objs)
# Sample five camera poses
for i in range(5):
    # Sample random camera location around the object
    location = bproc.sampler.sphere([0, 0, 0], radius=2, mode="SURFACE")
    # Compute rotation based on vector going from location towards poi
    rotation_matrix = bproc.camera.rotation_from_forward_vec(poi - location)
    # Add homog cam pose based on location an rotation
    cam2world_matrix = bproc.math.build_transformation_mat(location, rotation_matrix)
    bproc.camera.add_camera_pose(cam2world_matrix)

We sample here five random camera poses, where the location is on a sphere with a radius of 2 around the object. Each cameras rotation is such that it looks directly at the object and the camera faces upwards in Z direction.