Does the dataset contain transparent object instances?

Objectron Dataset

Objectron is a dataset of short object centric video clips with pose annotations.

Website • Dataset Format • Tutorials • License

The Objectron dataset is a collection of short, object-centric video clips, which are accompanied by AR session metadata that includes camera poses, sparse point-clouds and characterization of the planar surfaces in the surrounding environment. In each video, the camera moves around the object, capturing it from different angles. The data also contain manually annotated 3D bounding boxes for each object, which describe the object’s position, orientation, and dimensions. The dataset consists of 15K annotated video clips supplemented with over 4M annotated images in the following categories: bikes, books, bottles, cameras, cereal boxes, chairs, cups, laptops, and shoes. In addition, to ensure geo-diversity, our dataset is collected from 10 countries across five continents. Along with the dataset, we are also sharing a 3D object detection solution for four categories of objects — shoes, chairs, mugs, and cameras. These models are trained using this dataset, and are released in MediaPipe, Google's open source framework for cross-platform customizable ML solutions for live and streaming media.

Key Features

15000 annotated videos and 4M annotated images
All samples include high-res images, object pose, camera pose, point-cloud, and surface planes.
Ready to use examples in various tf.record formats, which can be used in Tensorflow/PyTorch.
Object-centric multi-views, observing the same object from different angles.
Accurate evaluation metrics, like 3D IoU for oriented 3D bounding boxes.

Dataset Format

The data is stored in the objectron bucket on Google Cloud storage. Check out the Download Data notebook for a quick review of how to download/access the dataset. The following assets are available:

The video sequences (located in /videos/class/batch-i/j/video.MOV files)
The annotation labels containing the 3D bounding boxes for objects. The annotation protobufs are located in /videos/class/batch-i/j/geometry.pbdata files. They are formatted using the object.proto. See [example] on how to parse the annotation files.
AR metadata (such as camera poses, point clouds, and planar surfaces). They are based on a_r_capture_metadata.proto. See example on how to parse these files.
Processed dataset: sharded and shuffled tf.records of the annotated frames, in tf.example format and videos in tf.SequenceExample format. These are used for creating the input data pipeline to your models. These files are located in /v1/records_shuffled/class/ and /v1/sequences/class/.
Supporting scripts to run evaluation based on the 3D IoU metric.
Supporting scripts to load the data into Tensorflow, Jax and Pytorch and visualize the dataset, including “Hello World” examples.
Supporting Apache Beam jobs to process the datasets on Google Cloud infrastructure.
The index of all available samples, as well as train/test splits for easy access and download.

Raw dataset size is 1.9TB (including videos and their annotations). Total dataset size is 4.4TB (including videos, records, sequences, etc.). This repository provides the required schemas and tools to parse the dataset.

class	bike	book	bottle	camera	cereal_box	chair	cup	laptop	shoe
#videos	476	2024	1928	815	1609	1943	2204	1473	2116
#frames	150k	576k	476k	233k	396k	488k	546k	485k	557k

Tutorials

License

Objectron is released under Computational Use of Data Agreement 1.0 (C-UDA-1.0). A copy of the license is available in this repository.

BibTeX

If you found this dataset useful, please cite our paper.

@article{objectron2021,
  title={Objectron: A Large Scale Dataset of Object-Centric Videos in the Wild with Pose Annotations},
  author={Adel Ahmadyan, Liangkai Zhang, Artsiom Ablavatski, Jianing Wei, Matthias Grundmann},
  journal={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  year={2021}
}

This is not an officially supported Google product. If you have any question, you can email us at [email protected] or join our mailing list at [email protected]

	for boxes, label, plane in zip(results, labels, planes):
	instances = label['2d_instance']
	instances_3d = label['3d_instance']
	visibilities = label['visibility']
	num_instances = 0
	for instance, instance_3d, visibility in zip(
	instances, instances_3d, visibilities):
	if (visibility > self._vis_thresh and
	self._is_visible(instance[0]) and instance_3d[0, 2] < 0):
	num_instances += 1
	# We don't have negative examples in evaluation.
	if num_instances == 0:
	continue

	iou_hit_miss = metrics.HitMiss(self._iou_thresholds)
	azimuth_hit_miss = metrics.HitMiss(self._azimuth_thresholds)
	polar_hit_miss = metrics.HitMiss(self._polar_thresholds)
	pixel_hit_miss = metrics.HitMiss(self._pixel_thresholds)

	num_matched = 0
	for box in boxes:
	box_point_2d, box_point_3d = box
	index = self.match_box(box_point_2d, instances, visibilities)
	if index >= 0:
	num_matched += 1
	pixel_error = self.evaluate_2d(box_point_2d, instances[index])

	# If you only compute the 3D bounding boxes from RGB images,
	# your 3D keypoints may be upto scale. However the ground truth
	# is at metric scale. There is a hack to re-scale your box using
	# the ground planes (assuming your box is sitting on the ground).
	# However many models learn to predict depths and scale correctly.
	#scale = self.compute_scale(box_point_3d, plane)
	#box_point_3d = box_point_3d * scale
	azimuth_error, polar_error, iou = self.evaluate_3d(box_point_3d, instances_3d[index])
	iou_hit_miss.record_hit_miss(iou)
	pixel_hit_miss.record_hit_miss(pixel_error, greater=False)
	azimuth_hit_miss.record_hit_miss(azimuth_error, greater=False)
	polar_hit_miss.record_hit_miss(polar_error, greater=False)

	if num_matched > 0:
	self._iou_ap.append(iou_hit_miss, num_instances)
	self._pixel_ap.append(pixel_hit_miss, num_instances)
	self._azimuth_ap.append(azimuth_hit_miss, num_instances)
	self._polar_ap.append(polar_hit_miss, num_instances)
	self._matched += num_matched

google-research-datasets / objectron Goto Github PK

objectron's Introduction

Objectron Dataset

Key Features

Dataset Format

Tutorials

License

BibTeX

objectron's People

Contributors

Stargazers

Watchers

Forkers

objectron's Issues

Recommend Projects

Recommend Topics

Recommend Org