Examples overview

We structure our examples into three different groups. We encourage everyone to start with the basic examples.

Each folder contains an example, some of those require external datasets and/or resources. Each example provides a valid configuration file(s) that can be used for getting some sort of output, a description, and, optionally, some resources.

Contents

• Basic Examples

• Advanced Examples

• Utility Examples

• BOP Dataset

• Integrated Datasets Examples

Basic Examples

Following examples will guide you through core functionality of BlenderProc. We recommend starting with them.

• basic example: Introduction. How to run a pipeline, and how, why and when certain things happen when running one.

• camera_sampling: Sampling of different camera positions inside of a shape with constraints for the rotation.

• light_sampling: Sampling of light poses inside of a geometrical shape.

• entity_manipulation: Changing various parameters of entities via selecting them in the config file.

• material_manipulation: Material selecting and manipulation.

• physics_positioning: Enabling simple simulated physical interactions between objects in the scene.

• semantic_segmentation: Generating semantic segmentation labels for a given scene.

• camera_object_pose: Load and render models given the intrinsics and extrinsics.

Advanced Examples

These examples introduce usage of advanced BlenderProc modules and/or of their combinations.

• auto_shading: How to change the shading mode of an object.

• camera_depth_of_field: Setting an object as the camera depth of field focus point.

• coco_annotations: Generating COCO annotations in polygon or RLE format.

• diffuse_color_image: How to render a scene without any lighting or shading.

• dust: How to add dust on top objects, to make them look more real.

• entity_displacement_modifier: Using displacement modifiers with different textures.

• lens_distortion: Add lens distortion from camera calibration to all output images.

• material_randomizer: Randomization of materials of selected objects.

• motion_blur_rolling_shutter: Generating motion blur and a rolling shutter effects.

• object_pose_sampling: Complex use of a 6D pose sampler.

• on_surface_object_sampling: Object pose sampling on a given surface surface.

• optical_flow: Obtaining forward/backward flow values between consecutive key frames.

• physics_convex_decomposition: This examples explains how to use a faster and more stable physics simulation (only linux)

• random_backgrounds: * Rendering an object in front of transparent background and then placing it on a random image

• random_room_constructor: Generating rooms and populating them with objects.

• stereo_matching: Compute distance image using stereo matching.

Utility Examples

This example is not a demonstration, but rather a tool to be used when developing your own pipeline.

• calibration: Verifying given camera intrinsics.

Benchmark for 6D Object Pose Estimation (BOP)

We provide example configs that interface with the BOP datasets.

• bop_challenge: Configs used to create the official synthetic data for the BOP challenge 2020

• bop_object_physics_positioning: Drop BOP objects on planes and randomize materials

• bop_object_on_surface_sampling: Sample upright poses on plane and randomize materials

• bop_scene_replication: Replicates whole scenes (object poses, camera intrinsics and extrinsics) from BOP

• bop_object_pose_sampling: Loads BOP objects and samples object, camera and light poses

Integrated Datasets Examples

We already support a lot of different datasets, check out the following examples, check out the following examples:

• blenderkit

• shapenet

• shapenet_with_scenenet

• shapenet_with_suncg

• scenenet

• scenenet_with_cctextures

• front_3d

• front_3d_with_improved_mat

• rock_essentials

• haven

• ikea

• pix3d

• AMASS_human_poses

• suncg_basic

• suncg_with_cam_sampling

• suncg_with_improved_mat

• suncg_with_object_replacer

• replica

• rock_essentials