Standardization of shipping container extraction from X-ray scan images using machine learning.

  • Category: BORDET
  • Swimlane: 2022-2023
  • Column: Assigned
  • Position: 4
  • Assignee: Olivier Debeir
  • Creator: Olivier Debeir
  • Started:
  • Created: 03/06/2022 15:33
  • Modified: 22/09/2022 14:14
  • Moved: 31/08/2022 16:50
Description

Description:

The general administration of customs and excise has at one's disposal different types of scanner to execute non-intrusive inspection of shipping containers.

Before running computer vision algorithms on the acquired X-ray scan, the container mounted on a truck or a trailer must be extracted from the image.

Algorithms were built to execute this task on some instruments but must be extended to others which have different geometry and calibration specificities.

The objective is to use machine learning to standardize the approach either by pre-processing images e.g. via style transfer and/or by adapting algorithms e.g. via Deep learning.

Required skills:

Python programming and basics of computer vision. Knowledge of deep learning is preferred.

Contact:

Sami Bali (sami.bali@minfin.fed.be), olivier.debeir@ulb.be

Sub-Tasks
Internal links
Comments
Olivier Debeir
Olivier Debeir Created at: 03/06/2022 15:37 Updated at: 03/06/2022 15:37

l’étudiant intéressé devra signer un Non Disclosure Agreement.

Olivier Debeir
Olivier Debeir Created at: 31/08/2022 16:50 Updated at: 31/08/2022 16:50

Bonjour Monsieur Meftah,

Ce que nous souhaitons faire est quelque peu « nouveau ». En effet, concernant le « style transfer », le but est d’harmoniser les images contenant des containers issus de différents scanners X-ray afin de n’utiliser qu’un seul set d’algorithmes dans la suite du pipeline de traitements. Nous n’avons rien trouvé dans la littérature scientifique qui abordait ce genre de procédé pour des image de containers. Par contre, certaines publications existent concernant l’harmonisation d’images médicales :

    Liu, M. et al. (2021). Style Transfer Using Generative Adversarial Networks for Multi-site MRI Harmonization. In: , et al. Medical Image Computing and Computer Assisted Intervention – MICCAI 2021. MICCAI 2021. Lecture Notes in Computer Science(), vol 12903. Springer, Cham. https://doi.org/10.1007/978-3-030-87199-4_30
    Enriquez Calzada, Patricia, Quantitative MR inter-scanner harmonization using image style transfer. http://resolver.tudelft.nl/uuid:4a41f267-a9fe-4e54-9dc7-fcd322b57ee5

Voici également d’autres articles liés au Style Transfer :

Reviews

· Yongcheng Jing, Yezhou Yang, Zunlei Feng, Jingwen Ye, Yizhou Yu, Mingli Song, Neural Style Transfer: A Review. arXiv:1705.04058. https://doi.org/10.48550/arXiv.1705.04058

· Yanghao Li, Naiyan Wang, Jiaying Liu, Xiaodi Hou, Demystifying Neural Style Transfer, arXiv:1701.01036. https://doi.org/10.48550/arXiv.1701.01036

Data augmentation

· Hernandez-Cruz , N., Cato, D. & Favela, J. Neural Style Transfer as Data Augmentation for Improving COVID-19 Diagnosis Classification. SN COMPUT. SCI. 2, 410 (2021). https://doi.org/10.1007/s42979-021-00795-2

· N. Mangaokar, J. Pu, P. Bhattacharya, C. K. Reddy and B. Viswanath, "Jekyll: Attacking Medical Image Diagnostics using Deep Generative Models," 2020 IEEE European Symposium on Security and Privacy (EuroS&P), 2020, pp. 139-157, https://doi.org/10.1109/EuroSP48549.2020.00017

More generic Style Transfer papers

· Ma, C., Huang, H., Sheffer, A., Kalogerakis, E. and Wang, R. (2014), Analogy-driven 3D style transfer. Computer Graphics Forum, 33: 175-184. https://doi.org/10.1111/cgf.12307

· Supriti Mulay, Keerthi Ram, Balamurali Murugesan, Mohanasankar Sivaprakasam, Style Transfer based Coronary Artery Segmentation in X-ray Angiogram. arXiv:2109.01441. https://doi.org/10.48550/arXiv.2109.01441

Comme je vous l’avais communiqué lors de notre entrevue, nous n’avons pas encore d’expertise en Style Transfer. Par contre, nos data scientists ont une connaissance des modèles en machine learning et réseaux de neurones artificiels tel que les auto-encoders ou réseaux de neurone convolutionnels.

Si vous avez d’autres questions, n’hésitez pas à nous en faire part.

Cordialement,

Sami Bali
Conseiller – Adviseur

SPF Finances | Douanes et Accises | Management des Risques | Team AI (techniques d’Analyse Innovantes & Artificial Intelligence)
FOD Financiën | Douane en Accijnzen | Risicomanagement | Team AI (Innovatieve analysetechnieken & Artificial Intelligence)

Boulevard du Roi Albert II, 33 bte 902 - 1030 Bruxelles
Koning Albert II- laan 33, bus 902, 1030 Brussel
Tél. : 0257 83168 | GSM : 0470835409