Skip to content

In this repo we use BlenderProc to create synthetic data for object detection.

Notifications You must be signed in to change notification settings

captain-bender/hangar-synthetic-data

Repository files navigation

blender_proc_experiments

In this repo we use BlenderProc to create synthetic data for object detection.

coco2yolo.py converts coco format to YOLO.

create_synthetic_data.py creates the synthetic data by performing renderings in the given scene using specific camera and moving specific objects.

filter_coco_annotations.py filters the categories to only the desired ones, since the models contains a big variety of category IDs.

verify_dataset.py verifies the yolo-transformed dataset

BlenderProc-related commands

blenderproc run create_synthetic_data.py
blenderproc vis hdf5 output/0.hdf5
blenderproc vis coco -i 0 -c coco_annotations.json -b output/coco_data

Required packages

The current repo has been tested in win 11, using BlenderProc 2.8.0

Author (or who to blame)

Angelos Plastropoulos

About

In this repo we use BlenderProc to create synthetic data for object detection.

Topics

Resources

Stars

Watchers

Forks

Languages