Skip to content

Latest commit

 

History

History
102 lines (73 loc) · 4 KB

README.md

File metadata and controls

102 lines (73 loc) · 4 KB

Replica dataset

This example introduces new tools for using replica dataset with BlenderProc.

Usage

Execute in the BlenderProc main directory:

blenderproc run examples/datasets/replica/main.py <path_to_the_replica_data_folder>  examples/datasets/replica/output
  • examples/datasets/replica/main.py: path to the python file with pipeline configuration.
  • <path_to_the_replica_data_folder>: Path to the replica dataset directory.
  • examples/datasets/replica/output: path to the output directory.

Visualization

Visualize the generated data:

blenderproc vis hdf5 example/replica/0.hdf5

Steps

  • Load the replica dataset : bproc.loader.load_replica().
  • Extract the floor from the room: bproc.object.extract_floor().
  • Find point of interest, all cam poses should look towards it: bproc.object.compute_poi().
  • Sample random camera location around the objects: bproc.sampler.sphere().
  • Adds camera pose to the scene: bproc.camera.add_camera_pose().
  • Enables normals and depth (rgb is enabled by default): bproc.renderer.enable_normals_output() bproc.renderer.enable_depth_output().
  • Renders all set camera poses: bproc.renderer.render().
  • Writes the output to .hdf5 containers: bproc.writer.write_hdf5()

Python file (main.py)

Global

# Load the replica dataset
objs = bproc.loader.load_replica(args.replica_data_folder, data_set_name="office_1", use_smooth_shading=True)

Note that "data_set_name": "office_1" is a replica room you want to render. This line can be replace with: "data_set_name": "<args:X>>", i.e. with an appropriate placeholder where X is a number of a placeholder.

As before all these values are stored in the GlobalStorage and are only used if no value are defined.

bproc.loader.load_replica handles importing objects from a given path. Here we are using smooth shading on all surfaces, instead of flat shading.

Floor extractor

# Extract the floor from the loaded room
floor = bproc.object.extract_floor(objs, new_name_for_object="floor")[0]
room = bproc.filter.one_by_attr(objs, "name", "mesh")

bproc.object.extract_floor() searches for the specified object and splits the surfaces which point upwards at a specified level away.

Replica camera sampler

# Init sampler for sampling locations inside the loaded replica room
point_sampler = bproc.sampler.ReplicaPointInRoomSampler(room, floor, height_list_values)

# define the camera intrinsics
bproc.camera.set_resolution(512, 512)

# Init bvh tree containing all mesh objects
bvh_tree = bproc.object.create_bvh_tree_multi_objects([room, floor])

poses = 0
tries = 0
while tries < 10000 and poses < 15:
    # Sample point inside room at 1.55m height
    location = point_sampler.sample(height=1.55)
    # Sample rotation (fix around X and Y axis)
    rotation = np.random.uniform([1.373401334, 0, 0], [1.373401334, 0, 2 * np.pi])
    cam2world_matrix = bproc.math.build_transformation_mat(location, rotation)

    # Check that obstacles are at least 1 meter away from the camera and have an average distance between 2 and 4 meters
    if bproc.camera.perform_obstacle_in_view_check(cam2world_matrix, {"min": 1.0, "avg": {"min": 2.0, "max": 4.0}}, bvh_tree):
        bproc.camera.add_camera_pose(cam2world_matrix)
        poses += 1
    tries += 1

This samples multiple camera poses per every imported room with camera-object collision check and obstacle check.

Material Manipulator

# Use vertex color of mesh as texture for all materials
for mat in room.get_materials():
    mat.map_vertex_color("Col", active_shading=False)

The mat.map_vertex_color() changes the material of the Replica objects so that the vertex color is renderer, this makes it possible to render colors on Replica scenes. Important: This does not mean that we load the complex texture files, we only use the low res vertex color for color rendering.

If you are in need of high-res color images, do we propose that you, yourself can try to implement the texture importer for the replica dataset.