A Lightweight Model for Satellite Pose

Estimation

Pierluigi Carcagnı̀(B) , Marco Leo , Paolo Spagnolo , Pier Luigi Mazzeo ,

and Cosimo Distante

CNR-ISASI, Ecotekne Campus via Monteroni snc, 73100 Lecce, Italy

pierluigi.carcagni@cnr.it

Abstract. In this work, a study on computer vision techniques for

automating rendezvous manoeuvres in space has been carried out. A
lightweight algorithm pipeline for achieving the 6 degrees of freedom
(DOF) object pose estimation, i.e. relative position and attitude, of a
spacecraft in a non-cooperative context using a monocular camera has
been studied. In particular, the considered lite architecture has been
never exploited for space operations and it allows to be compliant with
operational constraints, in terms of payload and power, of small satellite
platforms. Experiments were performed on a benchmark Satellite Pose
Estimation Dataset of synthetic and real spacecraft imageries specifi-
cally introduced for the challenging task of the 6DOF object pose esti-
mation in space. Extensive comparisons with existing approaches are
provided both in terms of reliability/accuracy and in terms of model size
that ineluctably affect resource requirements for deployment on space
vehicles.

Keywords: Spacecraft pose estimation · 6DOF pose · Deep learning ·

Monocular vision · Space imagery

1 Introduction
Estimating the 6 degrees of freedom (6DOF) pose of space-borne objects (e.g.,
satellites, spacecraft, orbital debris) is a crucial step in many space operations
such as docking, non-cooperative proximity tasks (e.g., debris removal), and
inter-spacecraft communications (e.g., establishing quantum links). It has unique
challenges that are not commonly encountered in the terrestrial setting. Due to
the importance of the problem, the Advanced Concepts Team (ACT) at ESA
recently held a benchmark competition called Kelvins Pose Estimation Challenge
(KPEC)1 .
The 6DOF pose estimation of objects in images is a traditional computer
vision task. Methods based on template matching [21] were initially used to this
aim. Unfortunately, when an object undergoes occlusions or drastic illumination
changes they become unreliable. To fix this issues local feature/keypoint match-
ing approaches were introduced [11]. The extraction of the fiducial points for
1
https://kelvins.esa.int/satellite-pose-estimation-challenge/home/.
c The Author(s), under exclusive license to Springer Nature Switzerland AG 2022
S. Sclaroff et al. (Eds.): ICIAP 2022, LNCS 13231, pp. 3–14, 2022.
https://doi.org/10.1007/978-3-031-06427-2_1
4 P. Carcagnı̀ et al.

the purpose of calculating the correspondences is carried out using handcrafted

descriptors such as Harris corners or Canny edges, lines e.g. by Hough trans-
form, or scale invariants such as SIFT, SURF and ORB [13]. The object poses
are subsequently obtained by solving a Perspective-n-Point (PnP) problem [10].
The above mentioned methods based on handcrafted descriptors, tend to pro-
duce low-quality outputs in difficult conditions (weak or absent texture of the
surfaces, strong lights, etc.) typical in an operational context in space.
To overcome these drawbacks, recent advances in pose estimation techniques,
for terrestrial applications, have been based on deep learning (DL) algorithms
instead. In general, these algorithms bypass the classic pipeline based on hand-
crafted features, and instead try, through the use of an appropriate deep convo-
lutional neural network (DCNN), to learn, in an end-to-end way, the non-linear
transformation between the two-dimensional space of the input image and the
six-dimensional exposure space of the network output. Learning is carried out
through appropriate supervised training.
Deep learning strategies have demonstrated robust behaviour under difficult
operating conditions in terms of scene illumination and object surface texture.
However, for efficient operations in terms of processed fps (frames per second)
it is necessary to have appropriate hardware with power constraints as opposed
to the classic approaches based on handcrafted descriptors.
Besides, they require a large amount of manual labels including the 2D key-
points, masks, 6D poses of objects, and other extra labels, which are usually
very costly. Many recent 6DOF pose estimation methods exploited 3D object
models to generate synthetic images for training because labels come for free.
However, due to the domain shift of data distributions between real images and
synthetic images, the network trained only on synthetic images fails to capture
robust features in real images for 6DOF pose estimation [19]. Another effective
pathway could be to combine the strength of deep neural networks and geo-
metric optimisation for example by incorporating a perspective-n-point (PnP)
solver in a deep neural architecture [3]. Most 6DOF pose estimation deep net-
works rely on an encoder-decoder architecture. To handle large scale variations
for 6D object pose estimation they can rely on an additional object detection
network or they exploit the inherent hierarchical architecture of the encoder
network, which extracts features at different scales [6]. Among different deep
architectures, the High-Resolution Network (HRNet) [15], initially introduced
for human pose estimation [17], has also recently gathered very relevant results
also for object detection and semantic segmentation. Differently from existing
state-of-the-art frameworks, that first encode the input image as a low-resolution
representation through a sub-network that is formed by connecting high-to-low
resolution convolutions in series (e.g., ResNet, VGGNet), and then recover the
high-resolution representation from the encoded low-resolution representation,
HRNet maintains high-resolution representations through the whole process. The
benefit is that the resulting representation is semantically richer and spatially
more precise.
 A Lightweight Model for Satellite Pose Estimation 5

This scientific fervour and the increased accessibility of space platforms, and
space data, recently pushed researchers to investigate the application of machine
learning researches in space activities. One of these ‘research frontiers’ concerns
the pose estimation of space objects for autonomous rendezvous or for capturing
uncontrolled targets in debris removal operations. Spacecraft (vehicles designed
for operation outside the earth’s atmosphere) and satellites (objects that orbit a
natural body) have two types of systems: payload, which comprises instruments
that facilitate the primary purpose of the spacecrafts; and operations systems,
which support the payload and allow it to reach, stay, and work in space. Modern
space compute systems are moving towards shared/re-configurable, multicore
systems which would be capable of running just lite DL models. Besides, another
limiting factor for on-board compute is power [9].
As a consequence, not all the DL models developed for terrestrial operations
can be exploited in space and then it is important to investigate how to find a
good trade-off between model complexity and accuracy/reliability.
In this work, a study on computer vision techniques for automating ren-
dezvous manoeuvres in space has been carried out. In particular, algorithms for
estimating the 6DOF object pose estimation, i.e. relative position and attitude,
of a spacecraft in a non-cooperative context using a monocular camera have been
studied. The term “non-cooperative” implies that the target spacecraft does not
have an active communication link or markers such as LEDs or reflectors useful
for distance and attitude estimation. To this aim, in this work, a 6DOF system
suited to be exploited on a Satellite Platform, in 50-kg micro-satellite class, has
been developed. The paper introduces and assesses, for the specific challenging
task of the 6DOf object pose estimation in space, a lite architecture inspired
by the Lite-HRNet [18] recently introduced for human pose estimation. By our
knowledge, this is the first attempt to exploit the lite-HRNet for space opera-
tions. Experiments were performed on the Spacecraft PosE Estimation Dataset
(SPEED) [14], the first publicly available machine learning set of synthetic and
real spacecraft imageries. Extensive comparisons with existing approaches are
provided both in terms of reliability/accuracy and in terms of model size that
ineluctably affect resource requirements for deployment on space vehicles [5].
The rest of the paper is organized as follows: Sect. 2 describes the problem, the
proposed algorithmic pipeline introduced to address it and the dataset of space
images used for experimental tests. Subsequently, Sect. 3 accurately describes
experimental results and it provides a deep discussion about the advisability
of the proposed trade-off between accuracy and computational requirement of
resources with respect to existing works in the literature. Finally, Sect. 4 con-
cludes the paper.

2 Methodology and Data

In this work, a PnP based pose estimation approach, exploiting keypoints

extracted by means of deep learning techniques, has been employed [16]. There
are extensive applications based on this approach in the literature, including
6 P. Carcagnı̀ et al.

action recognition, human-computer interaction, intelligent photo editing, pedes-

trian monitoring, etc. [7]. In particular, it is possible to divide the problem into
two main categories: top-down methods and bottom-up methods. According to
the top-down paradigm, the object of which to estimate the pose is first detected
and then its pose is estimated by exploiting only the informative content of the
region into the bounding box surrounding the identified object. The bottom-up
paradigm directly regresses the positions of the keypoints belonging to the same
object, or it detects all the keypoints in the scene and subsequently it groups the
keypoints on the same object. Although the top-down paradigm is more expen-
sive from a computational point of view, since in addition to the extraction of
keypoints it requires a preliminary phase of detection of the object of interest,
it is more accurate than the bottom-up paradigm [4].

2.1 Processing Pipeline

In this paper, a top-down paradigm has been exploited. The starting point was
the pipeline introduced in [2] and depicted in Fig. 1. The algorithmic components
of the pipeline have been modified in order to make the entire pipeline suitable
for use with embedded systems, and then to be compliant with computational
constraints for in space operations where power and resources are much more
limited than for terrestrial tasks. The pipeline is model-based, i.e. it relies on
the availability of a 3D model of the target object and the pose of the on-
board camera has to be estimated with respect to the actual model configuration
extracted from the acquired images. Hence, the pipeline consists of three main
processing modules performing:

1. the detection of the object in the image;

2. the keypoints estimation in the detected bounding-box of the object;
3. the computation of the 2D-3D keypoints correspondences, i.e. between the
available 3D points of the 3D model of the target and the estimated 2D
ones of the detected object, and final pose estimation by means of a PnP
(Perspective-n-Point) based algorithm.

Taking into account the limited computational resources and energy con-
sumption constraints on board of a spacecraft/satellite, the underlying idea of
this work is to address the keypoints extraction task of the target vehicle by
a ‘lite’ deep convolutional network in order to reduce the computational com-
plexity and the size of the trained model, in terms of parameters and memory
occupation.
In this paper, a Lite-HRNet architecture [18] has been implemented and
tested for the keypoints detection task, lowering this way the architectural
complexity, compared to HRNet based approach exploited in [2], and therefore
achieving the results of making the pipeline compliant with the computational
resources available on board of the vehicle.
The Lite-HRNet has been built-up following the same architectural strategies
exploited for HRNet. Starting from a high-resolution convolutional subnetwork
 A Lightweight Model for Satellite Pose Estimation 7

Fig. 1. Pose estimation pipeline.

as the first stage, high-to-low resolution subnetworks are gradually added and
connected in parallel, keeping the high-resolution of the initial image through
the whole process (Fig. 2).
Complexity is reduced by applying the efficient shuffle block, introduced
with ShuffleNet [20], to the HRNet architecture and then using fewer layers
and smaller widths. Moreover, the costly point-wise 1 × 1 convolution operation,
heavily used in the original shuffle blocks, has been replaced by a lightweight
unit, named conditional channel weighting. This allows, exploiting element-wise
weighting operations, the architecture to obtain a linear complexity with respect
to the number of channels instead of the quadratic complexity of the 1×1 convo-
lution operations in the original implementation of the shuffle block. Finally, two
lightweight functions have been introduced, a cross-resolution weighting function
and a spatial weighting function, in order to compute the weight maps from all
the channels across resolutions and for each resolution respectively compensating
the role played by the point-wise 1 × 1 convolution operation (Fig. 3).

Fig. 2. Starting from an high-resolution convolutional subnetwork as first stage, high-

to-low resolution subnetworks are gradually added and connected in parallel maintain-
ing high-resolution through the whole process.
8 P. Carcagnı̀ et al.

Fig. 3. (a) The shuffle block. (b) Conditional channel weighting block. H denotes the
cross-resolution weighting function, F denotes the spatial weighting function. Weights
distribution and representations from and to other resolutions are represented by the
dotted lines.

In this work, in particular, the Lite-HRNet-18 implementation, where suffix

18 indicates the number of layers, has been exploited. A size of 768 × 768 pixels
(the same of the solution in [2]) has been set for the input window and for the
output heatmaps.

2.2 The Dataset

For training and validating the proposed pipeline, the SPEED (Spacecraft Pose
Estimation Dataset) dataset [8] was exploited. It consists of 8-bit monochrome
images in JPEG format with resolution 1920×1200 pixels. The dataset has three
main folders of images: a folder containing 12000 synthetic images for training,
a folder with 2998 similar synthetic images for testing and a folder with 300
real images of the Tango satellite mock-up, same format and resolution as the
synthetic images. Ground truth data, in terms of 6DOF poses (position and
orientation), is provided only for the images in the training set. Some of the
training images are shown in Fig. 4. Images are particularly challenging due to
the large variations for the satellite in terms of lighting condition, distance from
the camera, orientation and background.
The camera model used for rendering the synthetic images is the same one
as the actual camera used for capturing the 300 images of the mock-up. The
related intrinsics camera parameters are: resolution = 1920 × 1200 pixels, focal
 A Lightweight Model for Satellite Pose Estimation 9

Fig. 4. Some Tango satellite images, at different distances from the camera, sampled
from the SPEED dataset.

length f = 17.6 mm, Pixel pitch p = 5.86 µm/px, Horizontal F oV = 35.452◦ ,

Vertical F oV = 22.595◦ .

3 Experimental Results
Algorithmic pipeline implementation and testing were carried out using the
Pytorch framework, and the Python language, in Ubuntu Linux 20.04 environ-
ment on a machine equipped with: Intel i7 processor; 64 GB of RAM; NVIDIA
TITATN RTX GPU card with 24 GB of RAM.
In the first experimental phase the 3D model of the Tango satellite mock-
up has been estimated (3D model and landmarks annotations are not provided
with SPEED dataset but only 6DOF ground truth for the training set). This
estimation has been carried out by picking up 9 close-up images from the training
folder. In each image, 11 keypoints were manually selected: they correspond to
some of the strongest visual characteristics in the images which, moreover, are
not occluded by other surfaces of the vehicle space.
Figure 5 shows the configurations of the 11 selected keypoints on 3 training
images.

Fig. 5. Three configurations of the selected keypoints for the initial estimation of the
3D model of the Tango satellite mock-up.
10 P. Carcagnı̀ et al.

Starting from the aforementioned 2D positions of the 11 keypoints in the

9 images and knowing the camera model parameters, the 3D structure of the
Tango satellite has been reconstructed by multi-view triangulation.
In particular, multi-view triangulation was performed by minimizing an
objective function


2

pi,j − πTj∗ (xi ) (1)
2
i,j

where pi;j denotes the 2D coordinates of the i-th landmark, obtained from the
j-th image, and xi the corresponding 3D landmark. Tj∗ is the ground truth pose
provided for the image j and πT is the projective transformation (6DOF pose and
intrinsic camera parameters are known) for the xi 3D landmark onto the image
plane. From Eq. 1 the 3D positions of the i selected landmarks were estimated
in each image j.
Exploiting the 6DOF ground truth data in the dataset and knowing the
estimated 3D model of the satellite, the ground truth 2D positions of landmarks
for all training images were obtained by projecting xi to the image plane by πT ∗ .
Finally, a bounding box was chosen in each image so that the 11 landmarks lie
in it.
The 2D positions of the landmarks and the corresponding bounding boxes
were then exploited to validate the proposed pipeline for 6DOF pose estimation.
The validation task was carried out by exploiting all available annotated
images (i.e. the images in the training folder of the SPEED dataset) and a
K-Fold Cross-Validation approach, with a number of folds equal to 6. In the
cross-validation, 5 folds were used for training the Lite-HRNet. The input of the
net were the patches obtained by cropping the original training image, around
the available bounding boxes surrounding the landmarks and the corresponding
2D positions of the 11 selected landmarks. They were then resized to a common
dimension of 768 × 768 pixels. In particular, the network was trained in order to
regress 11 heatmaps, of the same size of the input patch, corresponding to the
11 selected landmarks. Ground truth heatmaps were generated as 2D normal
distributions with mean equal to the ground truth 2D position of each landmark
and standard deviation of 1 pixel.
The remaining fold was then used for validating the capability of the net to
automatically estimate landmarks positions in unseen patches. This process was
repeated 6 times changing the fold used for validation among the available ones.
The object poses in all the patches extracted for the 6 validation folds (12.000
images) were then obtained by solving a Perspective-n-Point (PnP) problem [10]
exploiting the 2D-3D correspondences between the 11 predicted landmarks and
the 3D structure model of the satellite mock-up.
Lite-HRNet was trained by scratch, for each of the 6 validation steps, employ-
ing the ADAM optimizer with starting learning rate = 0.001 (dropped by a 0.1
factor at the 120th and 170th epochs respectively), momentum = 0.9 and weight
decay = 0.0001 parameters. A total number of 180 training epochs has been
chosen.
 A Lightweight Model for Satellite Pose Estimation 11

A rotation and a translation error were finally computed. In particular, indi-

cating with q ∗ and q the rotation quaternion ground truth and the estimated
one, the rotation error Er is defined as:

ER = 2 · cos−1 |q · q ∗ |, (2)

and the translation error is defined as:

t − t∗ 2
ET = , (3)
t2

for an overall error:

E = ER + ET . (4)
The experienced mean errors on validation folds were E R = 0.0302, E T =
0.0075 and E = 0.377 respectively. The E error is plotted in Fig. 6 versus the
distance between the camera and the target. In particular, the figure reports the
mean total error E on the y-axis, whereas the x-axis indicates distance ranges
from the camera, expressed in meters, of the corresponding detected TANGO
vehicle. The green bars correspond are related to the proposed pipeline whereas
the orange ones correspond to the error achieved by the pipeline in [2], where the
non-lite version of the HRNet was exploited. For a fair comparison, the pipeline
in [2] was applied in the identical operating conditions as for the proposed algo-
rithmic pipeline, i.e. without the final iterative refinement, in order to not add
external bias in the evaluation of the benefit in using the lightweight version of
the HRNet. As expected, the error decreased while the distance increased. It is
worth noting that for short distances errors for the two pipelines are compara-
ble. Of course, the gap between the performance of the lite version of HRNet
increased at long distances since, on small targets, the landmark positioning
failed.

Fig. 6. Mean total error vs Distance of the target from the camera.
12 P. Carcagnı̀ et al.

The numerical comparison between the proposed lite pipeline and the original
one is reported in Table 1. The overall rotational and translation errors increased
with a total error increasing from 0.0185 to 0.377. The rightmost column indi-
cates the model size, in terms of number of network parameters (M), for both
pipelines: it is worth noting that the original model had a size of 28.5 M whereas
the proposed one has a size of 1.1 M (more than 25 times less).

Table 1. Errors and complexity comparison.

Approach Mean rot. error Mean trans. error Mean total error Network #Params
[2] 0.0141 0.0044 0.0185 28.5 M
Proposed 0.0302 0.0075 0.0377 1.1 M

Another relevant aspect to take into account is the average prediction time.
The average time to predict the pose of the target vehicle by using the proposed
pipeline is about 5 ms (i.e. 5.10 ms) whereas it is about 12 ms (i.e. 11.94 ms) by
using the pipeline in [2].
Since, as largely stated in the introductory section, the computational load
and the hardware resources are crucial points for space applications, the above
numerical outcomes demonstrated that the introduce pipeline is particularly
suited for the rendezvous process since it relies on a model with much fewer
parameters allowing shorter computational time per image. To this aim, it is
worth noting that a typical rendezvous process can be divided into several phases,
including phasing of close-range rendezvous, final approaching, and docking.
Relative navigation and control are mainly used in the close-range rendezvous
phase, and orbit and attitude combined six-degree of freedom (6DOF) control
is used in the final approaching phase [12]. Under this perspective, it is worth
noting that, for close-range operations (up to about 20 m), the errors in pose
estimations obtained with the Lite-HRNet are comparable with those gathered
by state-of-the-art approaches relying on deep learning-based models.

4 Conclusions

This work has proven how the Lite-HRNet can be effectively exploited for
6DOF pose estimation for in space rendez-vous maneuvers. It represents the
first attempt of using this recent full-resolution architecture for in space opera-
tions. Experimental validations on a benchmark dataset demonstrated that, for
close range operations (up to 20 m), the errors in pose estimations are compara-
ble with those gathered by state-of-the-art approaches relying on deep learning
based models much more complex allowing, also, a half time of processing per
image. Future works will deal with building a top-down pipeline that, relying
on temporal/spatial filtering tricks, can alleviate also the computational load of
the object detection step. Besides, a tool for generating synthetic datasets of
 A Lightweight Model for Satellite Pose Estimation 13

photorealistic GAN-Generated scenes [1] will be introduced in order to train the

pipeline for any target.

Acknowledgement. This work was supported in part by the Ministry of Education,

University and Research under the grant PM3 AER01 01181 Modular Multi-Mission
Platform.

References
1. Brock, A., Donahue, J., Simonyan, K.: Large scale gan training for high fidelity
natural image synthesis. arXiv preprint arXiv:1809.11096 (2018)
2. Chen, B., Cao, J., Parra, A., Chin, T.J.: Satellite pose estimation with deep land-
mark regression and nonlinear pose refinement. In: Proceedings of the IEEE/CVF
International Conference on Computer Vision Workshops (2019)
3. Chen, B., Parra, A., Cao, J., Li, N., Chin, T.J.: End-to-end learnable geometric
vision by backpropagating PNP optimization. In: Proceedings of the IEEE/CVF
Conference on Computer Vision and Pattern Recognition, pp. 8100–8109 (2020)
4. Cheng, Y., Wang, B., Yang, B., Tan, R.T.: Monocular 3d multi-person pose esti-
mation by integrating top-down and bottom-up networks. In: Proceedings of the
IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 7649–
7659 (2021)
5. Hu, X., Chu, L., Pei, J., Liu, W., Bian, J.: Model complexity of deep learning: A
survey. arXiv preprint arXiv:2103.05127 (2021)
6. Hu, Y., Speierer, S., Jakob, W., Fua, P., Salzmann, M.: Wide-depth-range 6d object
pose estimation in space. In: Proceedings of the IEEE/CVF Conference on Com-
puter Vision and Pattern Recognition, pp. 15870–15879 (2021)
7. Jeon, M.H., Kim, A.: Prima6d: rotational primitive reconstruction for enhanced
and robust 6d pose estimation. IEEE Robot. Autom. Lett. 5(3), 4955–4962 (2020)
8. Kisantal, M., Sharma, S., Park, T.H., Izzo, D., Märtens, M., D’Amico, S.: Satellite
pose estimation challenge: Dataset, competition design, and results. IEEE Trans.
Aerosp. Electron. Syst. 56(5), 4083–4098 (2020)
9. Kothari, V., Liberis, E., Lane, N.D.: The final frontier: deep learning in space. In:
Proceedings of the 21st International Workshop on Mobile Computing Systems
and Applications, pp. 45–49 (2020)
10. Lepetit, V., Moreno-Noguer, F., Fua, P.: Epnp: an accurate o (n) solution to the
pnp problem. Int. J. Comput. Vision 81(2), 155 (2009)
11. Li, Z., Wang, G., Ji, X.: Cdpn: coordinates-based disentangled pose network for
real-time rgb-based 6-dof object pose estimation. In: Proceedings of the IEEE/CVF
International Conference on Computer Vision, pp. 7678–7687 (2019)
12. Luo, Y., Zhang, J., Tang, G.: Survey of orbital dynamics and control of space
rendezvous. Chin. J. Aeronaut. 27(1), 1–11 (2014)
13. Ma, J., Jiang, X., Fan, A., Jiang, J., Yan, J.: Image matching from handcrafted to
deep features: A survey. Int. J. Comput. Vision 129(1), 23–79 (2021)
14. Sharma, S., D’Amico, S.: Neural network-based pose estimation for noncoopera-
tive spacecraft rendezvous. IEEE Trans. Aerosp. Electron. Syst. 56(6), 4638–4658
(2020)
15. Sun, K., Xiao, B., Liu, D., Wang, J.: Deep high-resolution representation learning
for human pose estimation. In: CVPR (2019)
14 P. Carcagnı̀ et al.

16. Sundermeyer, M., Marton, Z.C., Durner, M., Brucker, M., Triebel, R.: Implicit 3d
orientation learning for 6d object detection from rgb images. In: Proceedings of
the European Conference on Computer Vision (ECCV), pp. 699–715 (2018)
17. Xiao, B., Wu, H., Wei, Y.: Simple baselines for human pose estimation and tracking.
In: European Conference on Computer Vision (ECCV) (2018)
18. Yu, C., Xiao, B., Gao, C., Yuan, L., Zhang, L., Sang, N., Wang, J.: Lite-hrnet: a
lightweight high-resolution network. In: Proceedings of the IEEE/CVF Conference
on Computer Vision and Pattern Recognition, pp. 10440–10450 (2021)
19. Zhang, S., Zhao, W., Guan, Z., Peng, X., Peng, J.: Keypoint-graph-driven learning
framework for object pose estimation. In: Proceedings of the IEEE/CVF Confer-
ence on Computer Vision and Pattern Recognition, pp. 1065–1073 (2021)
20. Zhang, X., Zhou, X., Lin, M., Sun, J.: Shufflenet: an extremely efficient convolu-
tional neural network for mobile devices. In: Proceedings of the IEEE Conference
on Computer Vision and Pattern Recognition, pp. 6848–6856 (2018)
21. Zhu, M., et al.: Single image 3d object detection and pose estimation for grasping.
In: 2014 IEEE International Conference on Robotics and Automation (ICRA), pp.
3936–3943. IEEE (2014)