We provide dimo_loader.py to easily load the dataset. This file is available on GitHub.
The dataset is structured using the BOP dataset format, with two additional files:
from dimo_loader import DimoLoader
from pathlib import Path
dimo_loader = DimoLoader()
dimo_path = Path('...') # Set path to downloaded dataset folder
dimo_ds = dimo_loader.load(dimo_path)
models = dimo_ds['models']
for model in models:
model_id = model['id'] # int
cad = model['cad'] # Path
# additional entries for size, diameter and
symmetries
real_jaigo_scenes = dimo_ds['real_jaigo']
for scene in real_jaigo_scenes:
scene_id = scene['id'] # str
for image in scene['images']:
img_id = image['id'] #
int
img_path = image['path'] #
str
camera = image['camera'] #
dict with intrinsics and extrinsics
scene_info = image['scene_info']
light = scene_info['light'] Lightmap id
carrier = scene_info['carrier'] Carrier id
parts = scene_info['parts']
Composition type
viewpoint = scene_info['viewpoint'] Viewpoint id
for obj in image['objects']:
obj_id = obj['id'] # int
model_2world =
obj['model_2world'] # Object to world, (4,4)
model_2cam =
obj['model_2cam'] # Object to camera, (4,4)
This work is
licensed under a Creative Commons
Attribution 4.0 International License.