Domain Randomization for Object Detection in Manufacturing Applications using Synthetic Data: A Comprehensive Study

Atsuto Maki; Duruo Li; Jacob Henningsson; Lars Hanson; M{\aa}rten Bj\"orkman; P\"ar M{\aa}rtensson; Xiaomeng Zhu

REVIEW 3 major objections 5 minor 32 references

Domain Randomization for Object Detection in Manufacturing Applications using Synthetic Data: A Comprehensive Study

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Domain-randomized synthetic data alone reaches 94-99.5% mAP@50 on real industrial parts.

desk verdict A useful synthetic-data benchmark and ablation study whose headline mAP numbers are likely optimistic because key DR choices were tuned on the small real test sets; still worth refereeing. read the letter →

arxiv 2506.07539 v1 pith:VOFPFSWS submitted 2025-06-09 cs.CV cs.AI

Xiaomeng Zhu , Jacob Henningsson , Duruo Li , Pär M{aa}rtensson , Lars Hanson , M{aa}rten Björkman , Atsuto Maki This is my paper · ORCID

classification cs.CVcs.AI

keywords domainrandomizationsim-to-realtransfersyntheticdatagenerationobjectdetectionmanufacturingYOLOv8pathtracingPBRmaterials

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Using only synthetic images rendered from CAD models, the authors train YOLOv8 detectors that reach 94.1% to 99.5% mAP@50 on real manufacturing scenes across four use cases. They argue that the key to this sim-to-real transfer is a domain randomization pipeline that varies object pose and texture, background, distractors, camera, lighting, and post-processing together, with path tracing and physically-based metal textures mattering most. The paper also introduces SIP15-OD, a benchmark of 15 industrial parts in three use cases, and releases the pipeline and dataset. If the claims hold, factories can build object detectors from CAD alone, skipping expensive manual annotation of real images.

What carries the argument

The central mechanism is a Blender-based synthetic data generation pipeline whose randomization spans five components: object quantity/pose/texture (with RGB, image, and PBR material options), background images from BG-20K plus distractor meshes, camera pose/FOV, area-light count/energy/color, and salt-and-pepper noise with Gaussian blur. The two load-bearing design choices are (1) rendering with Cycles path tracing rather than Eevee rasterization, which more faithfully simulates light reflection on metallic surfaces, and (2) physically-based rendering (PBR) metal textures for metallic parts, both of which narrow the visual gap between synthetic and real reflective objects. The pipeline automates segmentation masks and YOLO-format bounding boxes, and Guided Domain Randomization (GDR) is applied to constrain parameters to realistic ranges based on real-data feedback.

What would settle it

Hold out a set of freshly captured real images of the same parts in the same scenarios, never used in tuning; if a YOLOv8 model trained with the reported per-use-case settings does not keep mAP@50 above roughly 90%, the claim of closely covering the real data distribution outside the tuned scenes is not supported.

Watch

Extended reading notes

Core claim

The central claim is that a domain randomization pipeline spanning all five DR components—object characteristics, background, illumination, camera settings, and post-processing—generates synthetic images that closely cover the real-data distribution for manufacturing object detection. Trained exclusively on these images, YOLOv8 achieves mAP@50 of 96.4% on the public robotics dataset, surpassing the earlier pipeline's 83.6% with the same detector family, and 94.1%, 99.5%, and 95.3% on the three SIP15-OD use cases. The ablation studies attribute the gains to path tracing over rasterization, PBR metal textures for metallic objects, and the consistent inclusion of post-processing and distractors. The authors further find that data size and epoch limits saturate, and that GDR-based tuning of randomized parameters (e.g., rotation limits) prevents unrealistic scenarios. To our knowledge, this is the first study to integrate all DR components and evaluate across multiple manufacturing datasets.

Load-bearing premise

The real test images used to tune the randomization settings are assumed to be representative of the deployment environment, even though those settings were selected based on performance on those same images.

Editorial extensions

If this is right

Manufacturers with CAD models but no labeled images can train detection models for quality inspection and robotic picking at above 94% mAP@50, removing the annotation bottleneck.
The factor hierarchy (path tracing over rasterization, PBR textures for metal, post-processing and distractors always beneficial) provides concrete configuration rules for future synthetic-data pipelines.
Data and epoch saturation means compute budgets can be planned: roughly 800 images per category for controlled scenes, 1.5k-6k for varied ones, and up to 1k-2.5k training epochs for best real-test performance.
The SIP15-OD benchmark offers a public test bed covering textureless, metallic, and visually similar industrial parts across multiple scenarios.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Because GDR configurations were tuned on the same real test images used for evaluation, the headline numbers likely overstate expected performance on truly unseen factory scenes; the relative ordering of DR factors is more transferable than the exact mAP values.
The small real test sets (190 and 321 images) imply wide confidence bounds, so a user should re-evaluate with their own held-out images before deploying.
A parsimonious extension of this work would test whether the identified factor ranking transfers to other object categories (e.g., plastic housings or fabric) by running the same ablations with PBR plastic and cloth materials.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: the paper's real contributions are the SIP15-OD dataset (15 industrial parts, 321 real images across three use cases) and the systematic ablation of DR components in a manufacturing context, including path tracing vs rasterization and PBR vs random textures. The synthetic-only results on the public Horvath robotics dataset (96.4 mAP@50 with YOLOv8, beating the published baseline by more than 5 points) are credible and a useful data point.

The methodology is straightforward and the experimental load is substantial: three repetitions, multiple use cases, and a public code/dataset link. The failure-case analysis is honest. This is not a sloppy paper; it is a careful empirical study with one real evaluation-protocol flaw.

That flaw: Section V-B and Table VI show the "ideal" data sizes, epochs, and rotation limits were selected using performance on the same real test images that produce the reported mAP. That is test-set tuning, and with test sets of 190 and 321 images (scenario subsets down to 27), the chosen configurations can overfit to those specific photos. The paper says each experiment was repeated three times but reports only averages, with no error bars, so the stability of the numbers is not assessable. The central claim that the pipeline "closely covers the distribution of real data" is therefore not established for unseen scenes, which is the deployment setting the conclusion targets. I agree with the stress-test note: the reported numbers should be read as optimistic upper bounds.

Two smaller issues. First, Table VI's "ideal" numbers are presented as insight, but because they are test-selected they are more like fitted hyperparameters; the qualitative conclusions (diminishing returns at larger data sizes, path tracing helps) survive, but the exact numbers do not. Second, reference [29] for BG-20K cites "J. Doe and J. Smith, Computer Vision Journal"—that is a placeholder citation and must be fixed before publication. Not a scientific flaw, but an editorial red flag.

Who is this for? People working on sim-to-real object detection in industrial settings. They get a new benchmark, a concrete pipeline, and a useful comparison of rendering choices. It deserves a serious referee: the empirical core is valuable, the dataset is public, and the flaws are fixable. I would accept it for review, but require the authors to report per-run results or error bars, disclose the GDR/test-set tuning procedure, and evaluate with a properly held-out validation split. The placeholder reference must be corrected.

Referee Report

3 major / 5 minor

Summary. The paper presents a synthetic data generation pipeline for industrial object detection that applies domain randomization across object characteristics, background, camera, illumination, and post-processing. The authors introduce a new dataset, SIP15-OD, containing 321 real images of 15 industrial objects from three manufacturing use cases, and also evaluate on the public robotics dataset of Horvath et al. Using YOLOv8 models trained exclusively on synthetic images, they report mAP@50 scores of 96.4% on the robotics dataset and 94.1%, 99.5%, and 95.3% on the three SIP15-OD use cases. They identify material properties, rendering methods, post-processing, and distractors as key factors, and release the dataset and code.

Significance. If the results hold, the paper is a useful practical demonstration that synthetic-only training can reach high detection accuracy in controlled manufacturing settings. The breadth of the study across multiple use cases and DR components, the public release of the dataset and code, and the careful ablations of rendering and material properties are clear strengths. However, the central claim that the pipeline 'closely covers the distribution of real data' is currently supported only by numbers obtained with configurations tuned on the same small real test sets, and without error bars. The significance would be considerably strengthened by an evaluation protocol that separates configuration selection from final evaluation.

major comments (3)

[V-B, Table VI] The best-performing configurations were selected using performance on the same real test images that are later used for evaluation. Section V-B states that GDR was used to 'selectively adjust parameters,' and Table VI explicitly reports 'ideal data sizes and training epochs' chosen by test performance. Because the real test sets are small (190 images for the robotics dataset; 321 total for SIP15-OD, with scenario subsets as small as 27 images), this protocol makes the reported mAP values optimistic upper bounds and does not establish the conclusion that the pipeline covers the real distribution for unseen scenes. The manuscript should either use a held-out test set for final evaluation, evaluate with fixed a priori configurations, or explicitly reframe the claims as applying to these specific test images.
[V, Tables IV-V] The paper states that each experiment was repeated three times and average results were reported, but no variance measures, per-run values, or significance tests are provided. Given the small test sets and the small differences between several configurations in Tables IV and V (for example, 89.6 vs 96.4 for path tracing versus rasterization in the Robotics row, and 92.0 vs 94.1 for U1), the claim that path tracing consistently outperforms rasterization and the other factor-ablation conclusions are not supported without confidence intervals or per-run results. Please report standard deviations or individual run values.
[IV, Table I; VI-C, Fig. 5] The real test sets are very small at the scenario and class level: S1 has 27 images per use case, and per-class object counts are far lower for several categories (e.g., hook, electricity12v, gear1). mAP estimates on such small samples have wide uncertainty, and the class-wise failure analysis in Section VI-C is consequently fragile. The authors should quantify this uncertainty (for example, with bootstrap confidence intervals) or validate on additional images before drawing strong conclusions about which object types are intrinsically hard.

minor comments (5)

[III-A] The text says 'Camera aimes at the center'; this should be 'aims.'
[References] Reference [29] is given as 'J. Doe and J. Smith' with a generic venue; this appears to be a placeholder and should be replaced with the actual BG-20K citation.
[III-A] The acronym PBR is used without expansion at first use; please define 'Physically Based Rendering' in the main text (it is only expanded in the appendix table).
[Table VI] The header 'Ideal Data size (per category) and training epoch' is ambiguous because Section V-B reports total numbers (7,500, 18,000, 10,500 images); clarify whether the table entries are per-category or total.
[Fig. 3 caption] The caption for Fig. 3(g) says 'objects in a blue delivery box (U2 as example)' but the surrounding text and other captions refer to U1-S2; please check and align the caption with the displayed images.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the reported mAP scores are empirical evaluations against real images, though the selection of some DR parameters on the same test images makes the scores optimistic.

full rationale

The paper is an empirical study rather than a derivation. Its central claim—that a synthetic-only domain-randomization pipeline can achieve high mAP on real industrial images—is supported by evaluations on an external public robotics dataset (Horvath et al.) and on the newly introduced SIP15-OD dataset. No equation in the paper defines the reported mAP in terms of the DR parameters, and no fitted parameter is renamed as a prediction. The two self-citations ([1] and [5]) appear only in motivational or general statements about sim-to-real difficulty and are not load-bearing for the reported results. The main methodological caveat is that some DR choices (rotation limits, textures, data sizes, and epochs; Section V-B and Table VI) were selected using the same real test images later used for evaluation, so the reported scores are likely optimistic; however, this is an evaluation-protocol concern rather than a circularity in the derivation, because the mAP values are not forced by construction and the robotics benchmark is external. No circular step is present.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central empirical claims rest on hand-tuned DR parameters (rotation limits, data sizes, epochs, texture choices) that were selected using the same real test sets used for final evaluation. No new theoretical entities are introduced.

free parameters (4)

U1 object rotation limit = 30 degrees in x and y
In Section V-B, U1 limits object rotation to 30 degrees after systematically identifying key factors based on real test performance; this hand-tuned constraint is part of the reported best configuration.
Per-use-case training data size = 7,500 (U1), 18,000 (U2), 10,500 (U3), 4k/8k (Robotics)
Table VI reports ideal data sizes selected by performance on real test data; these are fitted parameters, not derived from a principle.
Per-use-case training epochs = 500 (Robotics), 1,000 or 2,500 (U1), 500/1,000 (U2), 1,000 (U3)
Table VI lists ideal training epochs that produced the best results; these are empirical fits to the test set.
Texture selection per use case = PBR metal for Robotics and U3, PBR metal/plastic for U1, PBR plastic for U2
Table IV best configurations are chosen by comparing on real test data; the choice is fitted, especially the switch to realistic textures.

assumptions (3)

domain assumption CAD models used for synthetic rendering are faithful geometric and appearance proxies for the real objects.
The entire sim-to-real pipeline assumes that the 3D models and materials used in Blender represent the real parts; if CAD models are inaccurate (for example missing chamfers or reflective properties), the synthetic training distribution will diverge from deployment. Invoked throughout Sections III and IV.
domain assumption Real test images in SIP15-OD and the Horvath dataset are representative of the deployment environment and are large enough to yield stable mAP estimates.
With 70 to 160 real images per use case and tens of images per class, the reported mAP values and ablation rank orders may have high variance; the paper assumes these measurements are reliable. Invoked in Section V-B and Tables III through V.
domain assumption Blender Cycles (path tracing) and EEVEE (rasterization) renderings capture the visual features that matter for CNN-based object detection.
The paper compares rendering engines and concludes path tracing is superior, but does not validate against real image statistics; it assumes renderer outputs are a valid proxy for real camera images. Discussed in Section VI-A.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Domain Randomization for Object Detection in Manufacturing Applications using Synthetic Data: A Comprehensive Study." pith.science (2026). https://pith.science/paper/VOFPFSWS

@misc{pith2026250607539,
  author       = {Pith},
  title        = {Pith review of: Domain Randomization for Object Detection in Manufacturing Applications using Synthetic Data: A Comprehensive Study},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VOFPFSWS}},
  note         = {Machine review of arXiv:2506.07539}
}

read the original abstract

This paper addresses key aspects of domain randomization in generating synthetic data for manufacturing object detection applications. To this end, we present a comprehensive data generation pipeline that reflects different factors: object characteristics, background, illumination, camera settings, and post-processing. We also introduce the Synthetic Industrial Parts Object Detection dataset (SIP15-OD) consisting of 15 objects from three industrial use cases under varying environments as a test bed for the study, while also employing an industrial dataset publicly available for robotic applications. In our experiments, we present more abundant results and insights into the feasibility as well as challenges of sim-to-real object detection. In particular, we identified material properties, rendering methods, post-processing, and distractors as important factors. Our method, leveraging these, achieves top performance on the public dataset with Yolov8 models trained exclusively on synthetic data; mAP@50 scores of 96.4% for the robotics dataset, and 94.1%, 99.5%, and 95.3% across three of the SIP15-OD use cases, respectively. The results showcase the effectiveness of the proposed domain randomization, potentially covering the distribution close to real data for the applications.

Figures

Figures reproduced from arXiv: 2506.07539 by the authors.

**Figure 1.** Overview of our data generation pipeline and evaluation method. Processes shown in blue text indicate where domain [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

**Figure 2.** 3D models of the SIP15-OD dataset. Their names [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** (a-d) Samples of synthetic images. (e-h) Samples of real images. (e) Robotics use cases [10]. (f) S1: objects on a [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (9 more)

**Figure 4.** Figure 4: Comparison of synthetic data rendered with different [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

**Figure 5.** Figure 5: Class-wise results (%) for each use case. All mAP@50 scores are shown in numbers, with scores below 90% [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

**Figure 6.** Figure 6: Samples from the SIP15-OD dataset U3 illustrate the synthetic data generation pipeline. (a) Shows the 3D scene [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

**Figure 7.** Figure 7: Sample images from the robotic dataset. (a-d) Synthetic images generated by our pipeline. (e-h) Real images provided [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

**Figure 8.** Figure 8: Samples images from the SIP15-OD dataset, U1. (a-d) Synthetic images. (e-h) Real images from S1. (i-l) Real images [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]

**Figure 9.** Figure 9: Sample images from the SIP15-OD dataset, U2. (a-d) Synthetic images. (e-h) Real images from S1. (i-l) Real images [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]

**Figure 10.** Figure 10: Sample images from the SIP15-OD dataset, U3. (a-d) Synthetic images. (e-h) Real images from S1. (i-l) Real [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]

**Figure 13.** Figure 13: Samples of the predictions for the SIP15-OD dataset, [PITH_FULL_IMAGE:figures/full_fig_p013_13.png]

**Figure 14.** Figure 14: Samples of the predictions for the SIP15-OD dataset, [PITH_FULL_IMAGE:figures/full_fig_p013_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 27 canonical work pages

[29]

Bg-20k: A diverse background dataset for computer vision applications,

J. Doe and J. Smith, “Bg-20k: A diverse background dataset for computer vision applications,”Computer Vision Journal, vol. 34, no. 2, pp. 123–135, 2021

work page 2021
[1]

Towards sim-to-real industrial parts classification with synthetic dataset,

X. Zhu, T. Bilal, P. M ˚artensson, L. Hanson, M. Bj ¨orkman, and A. Maki, “Towards sim-to-real industrial parts classification with synthetic dataset,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2023, pp. 4453–4462

work page 2023
[2]

Deep learning methods for object detection in smart manufacturing: A survey,

H. M. Ahmad and A. Rahimi, “Deep learning methods for object detection in smart manufacturing: A survey,”Journal of Manufacturing Systems, vol. 64, pp. 181–196, 2022

work page 2022
[3]

Yolo-based object detec- tion models: A review and its applications,

A. Vijayakumar and S. Vairavasundaram, “Yolo-based object detec- tion models: A review and its applications,”Multimedia Tools and Applications, pp. 1–40, 2024

work page 2024
[4]

Generating images with physics- based rendering for an industrial object detection task: Realism versus domain randomization,

L. Eversberg and J. Lambrecht, “Generating images with physics- based rendering for an industrial object detection task: Realism versus domain randomization,”Sensors, vol. 21, no. 23, p. 7901, 2021

work page 2021
[5]

Automated assembly quality inspection by deep learning with 2d and 3d synthetic cad data,

X. Zhu, P. M ˚artensson, L. Hanson, M. Bj ¨orkman, and A. Maki, “Automated assembly quality inspection by deep learning with 2d and 3d synthetic cad data,”Journal of Intelligent Manufacturing, pp. 1–16, 2024

work page 2024
[6]

Domain randomization for transferring deep neural networks from simulation to the real world,

J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel, “Domain randomization for transferring deep neural networks from simulation to the real world,” in2017 IEEE/RSJ international con- ference on intelligent robots and systems (IROS). IEEE, 2017, pp. 23–30

2017
[7]

CAD-based Learning for Egocentric Object Detection in Industrial Context,

J. Cohen, C. F. Crispim-Junior, C. Grange-Faivre, and L. Tougne, “CAD-based Learning for Egocentric Object Detection in Industrial Context,” in15th International Conference on Computer Vision Theory and Applications, vol. 5. Valletta, Malta: SCITEPRESS - Science and Technology Publications, Feb. 2020, pp. 644–651. [Online]. Available: https://hal.science/...

work page 2020

Show all 32 references

[8]

Two-stage filtering method to improve the performance of object detection trained by synthetic dataset in heavily cluttered industry scenes,

P. Tang, Y . Guo, G. Zheng, L. Zheng, J. Pu, J. Wang, and Z. Chen, “Two-stage filtering method to improve the performance of object detection trained by synthetic dataset in heavily cluttered industry scenes,”The Visual Computer, vol. 40, no. 3, pp. 2015–2034, 2024

2015
[9]

A novel method for object detection using deep learning and cad models,

I. G. B. Sampaio, L. Machaca, J. Viterbo, and J. Gu ´erin, “A novel method for object detection using deep learning and cad models,” arXiv preprint arXiv:2102.06729, 2021

2021 arXiv
[10]

Object detection using sim2real domain randomization for robotic applica- tions,

D. Horv ´ath, G. Erd ˝os, Z. Istenes, T. Horv ´ath, and S. F ¨oldi, “Object detection using sim2real domain randomization for robotic applica- tions,”IEEE Transactions on Robotics, vol. 39, no. 2, pp. 1225–1243, 2022

2022
[11]

Towards fully-synthetic training for industrial applications,

C. Mayershofer, T. Ge, and J. Fottner, “Towards fully-synthetic training for industrial applications,” inLISS 2020: Proceedings of the 10th International Conference on Logistics, Informatics and Service Sciences. Springer Singapore, 2021, pp. 765–782

2020
[12]

Training deep networks with synthetic data: Bridging the reality gap by domain randomization,

J. Tremblay, A. Prakash, D. Acuna, M. Brophy, V . Jampani, C. Anil, others, and S. Birchfield, “Training deep networks with synthetic data: Bridging the reality gap by domain randomization,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Worksh...

2018
[13]

Structured domain random- ization: Bridging the reality gap by context-aware synthetic data,

A. Prakash, S. Boochoon, M. Brophy, D. Acuna, E. Cameracci, G. State, O. Shapira, and S. Birchfield, “Structured domain random- ization: Bridging the reality gap by context-aware synthetic data,” in 2019 International Conference on Robotics and Automation (ICRA). Montreal, QC,...

2019
[14]

Efficientdet: Scalable and effi- cient object detection,

M. Tan, R. Pang, and Q. V . Le, “Efficientdet: Scalable and effi- cient object detection,”Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 10 781–10 790, 2020

2020
[15]

Ultralytics YOLO,

G. Jocher, A. Chaurasia, and J. Qiu, “Ultralytics YOLO,” Jan. 2023. [Online]. Available: https://github.com/ultralytics/ultralytics

2023
[16]

End-to-end object detection with transformers,

N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, and S. Zagoruyko, “End-to-end object detection with transformers,”Pro- ceedings of the European Conference on Computer Vision (ECCV), pp. 213–229, 2020

2020
[17]

A review on anchor assignment and sampling heuristics in deep learning-based object detection,

X.-T. V o and K.-H. Jo, “A review on anchor assignment and sampling heuristics in deep learning-based object detection,”Neurocomputing, vol. 506, pp. 96–116, 2022

2022
[18]

A survey of modern deep learning based object detection models,

S. S. A. Zaidi, M. S. Ansari, A. Aslam, N. Kanwal, M. Asghar, and B. Lee, “A survey of modern deep learning based object detection models,”Digital Signal Processing, vol. 126, p. 103514, 2022

2022
[19]

Mujoco: A physics engine for model-based control,

E. Todorov, T. Erez, and Y . Tassa, “Mujoco: A physics engine for model-based control,”2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 5026–5033, 2012

2012
[20]

Very deep convolutional networks for large-scale image recognition,

K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,”arXiv preprint arXiv:1409.1556, 2014

2014 arXiv
[21]

Shreiner, G

D. Shreiner, G. Sellers, J. Kessenich, and B. Licea-Kane,OpenGL Programming Guide: The Official Guide to Learning OpenGL, Version 4.3. Addison-Wesley, 2013

2013
[22]

Faster r-cnn: Towards real- time object detection with region proposal networks,

S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real- time object detection with region proposal networks,” inAdvances in neural information processing systems, 2015, pp. 91–99

2015
[23]

Blender - a 3d modelling and rendering package,

Blender Online Community, “Blender - a 3d modelling and rendering package,” 2021, blender Foundation, Blender Institute. [Online]. Available: http://www.blender.org

2021
[24]

A. S. Glassner,An introduction to ray tracing. Morgan Kaufmann, 1989

1989
[25]

Pybullet, a python module for physics simulation for games, robotics and machine learning,

E. Coumans and Y . Bai, “Pybullet, a python module for physics simulation for games, robotics and machine learning,” 2016. [Online]. Available: http://pybullet.org

2016
[26]

Yolov4: Op- timal speed and accuracy of object detection,

A. Bochkovskiy, C.-Y . Wang, and H.-Y . M. Liao, “Yolov4: Op- timal speed and accuracy of object detection,”arXiv preprint arXiv:2004.10934, 2020

2004 arXiv
[27]

Framing image descrip- tion as a ranking task: Data, models and evaluation metrics,

M. Hodosh, P. Young, and J. Hockenmaier, “Framing image descrip- tion as a ranking task: Data, models and evaluation metrics,”Journal of Artificial Intelligence Research, vol. 47, pp. 853–899, 2013

2013
[28]

CC Texture Dataset,

L. Demes, “CC Texture Dataset,” https://ambientcg.com/, 2023, ac- cessed: 2017-02-01

2023
[30]

[Online]

Blender Foundation,Blender Python API, 2021, accessed: 2021-09-10. [Online]. Available: https://docs.blender.org/api/current/

2021
[31]

[Online]

——,Blender Manual: Cycles, 2021, accessed: 2021-07-05. [Online]. Available: https://docs.blender.org/manual/en/latest/render/ cycles/index.html

2021
[32]

[Online]

——,Blender Manual: Eevee, 2021, accessed: 2021-07-05. [Online]. Available: https://docs.blender.org/manual/en/latest/render/ eevee/index.html

2021

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [29]

Bg-20k: A diverse background dataset for computer vision applications,

J. Doe and J. Smith, “Bg-20k: A diverse background dataset for computer vision applications,”Computer Vision Journal, vol. 34, no. 2, pp. 123–135, 2021

work page 2021

[2] [1]

Towards sim-to-real industrial parts classification with synthetic dataset,

X. Zhu, T. Bilal, P. M ˚artensson, L. Hanson, M. Bj ¨orkman, and A. Maki, “Towards sim-to-real industrial parts classification with synthetic dataset,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2023, pp. 4453–4462

work page 2023

[3] [2]

Deep learning methods for object detection in smart manufacturing: A survey,

H. M. Ahmad and A. Rahimi, “Deep learning methods for object detection in smart manufacturing: A survey,”Journal of Manufacturing Systems, vol. 64, pp. 181–196, 2022

work page 2022

[4] [3]

Yolo-based object detec- tion models: A review and its applications,

A. Vijayakumar and S. Vairavasundaram, “Yolo-based object detec- tion models: A review and its applications,”Multimedia Tools and Applications, pp. 1–40, 2024

work page 2024

[5] [4]

Generating images with physics- based rendering for an industrial object detection task: Realism versus domain randomization,

L. Eversberg and J. Lambrecht, “Generating images with physics- based rendering for an industrial object detection task: Realism versus domain randomization,”Sensors, vol. 21, no. 23, p. 7901, 2021

work page 2021

[6] [5]

Automated assembly quality inspection by deep learning with 2d and 3d synthetic cad data,

X. Zhu, P. M ˚artensson, L. Hanson, M. Bj ¨orkman, and A. Maki, “Automated assembly quality inspection by deep learning with 2d and 3d synthetic cad data,”Journal of Intelligent Manufacturing, pp. 1–16, 2024

work page 2024

[7] [6]

Domain randomization for transferring deep neural networks from simulation to the real world,

J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel, “Domain randomization for transferring deep neural networks from simulation to the real world,” in2017 IEEE/RSJ international con- ference on intelligent robots and systems (IROS). IEEE, 2017, pp. 23–30

2017

[8] [7]

CAD-based Learning for Egocentric Object Detection in Industrial Context,

J. Cohen, C. F. Crispim-Junior, C. Grange-Faivre, and L. Tougne, “CAD-based Learning for Egocentric Object Detection in Industrial Context,” in15th International Conference on Computer Vision Theory and Applications, vol. 5. Valletta, Malta: SCITEPRESS - Science and Technology Publications, Feb. 2020, pp. 644–651. [Online]. Available: https://hal.science/...

work page 2020