Abstract
In practical industrial visual inspection tasks, foreign object data are difficult to collect and accumulate, hence few-shot object detection has gradually become the focus of research. It has been observed that industrial foreign objects are often different from natural data and are always fractal objects. Its form is a rough or fragmented geometric shape, and its features are relatively monotonous and difficult to distinguish. Optimization-based meta-learning is a powerful approach to few-shot learning. It updates model weights through a parameter optimization strategy enabling more efficient learning when faced with new tasks with few samples. Therefore, we proposed a gradient scout strategy, which used the intelligent optimization idea to optimize the meta-training outer-loop parallel gradient optimization method to improve the training effect of few-shot fractal object detection. Meanwhile, we proposed a fractal information amplified learning module, which could improve the detection ability of few-shot fractal objects more quickly under the same training period. They formed FLGS (fractal information amplified learning with gradient scout), which was deployed at zero cost. YOLOv7 was advanced to a new industrial fractal object detection model under FLGS. The experimental results on the IGBT surface foreign object dataset showed that our gradient scout strategy was superior to the other eight few-shot meta-learning algorithms. FLGS significantly accelerated the improvement of fractal object detection ability and maintained a high-level mean average precision.
Similar content being viewed by others
Data availability
The datasets generated during the current study are available from the corresponding author on reasonable request.
References
Bochkovskiy A, Wang C, Liao H (2020) Yolov4: optimal speed and accuracy of object detection. Preprint at arXiv:2004.10934
Chen H, Wang Y, Wang G et al (2018) LSTD: a low-shot transfer detector for object detection. In: AAAI conference on artificial intelligence. AAAI, New Orleans, LA, pp 2836–2843
Chopra S, Hadsell R, LeCun Y (2005) Learning a similarity metric discriminatively, with application to face verification. In: IEEE conference on computer vision and pattern recognition, vol 1. IEEE, San Diego, pp 539–546
Fan Q, Zhuo W, Tang C et al (2020) Few-shot object detection with attention-RPN and multi-relation detector. In: IEEE conference on computer vision and pattern recognition. IEEE, ELECTR NETWORK, pp 4012–4021s
Finn C, Abbeel P, Levine S (2017) Model-agnostic meta-learning for fast adaptation of deep networks. Preprint at arXiv:1703.03400
Fu G, Sun P, Zhu W et al (2019) A deep-learning-based approach for fast and robust steel surface defects classification. Opt Lasers Eng 121:397–405
Girshick R, Donahue J, Darrell T et al (2014) Rich feature hierarchies for accurate object detection and semantic segmentation. In: IEEE conference on computer vision and pattern recognition. IEEE, Columbus, OH, pp 580–587
Gu K, Zhang Y, Qiao J (2021) Ensemble meta-learning for few-shot soot density recognition. IEEE Trans Ind Inf 13(3):2261–2270
He K, Gkioxari G, Dollár P et al (2017) Mask R-CNN. IEEE Trans Pattern Anal Mach Intell 42:2961–2969
Hoffer E, Ailon N (2014) Deep metric learning using triplet network. Preprint at arXiv:1412.6622
Hospedales T, Antoniou A, Micaelli P et al (2022) Meta-learning in neural networks: a survey. IEEE Trans Pattern Anal Mach Intell 44(9):5149–5169
Hu H, Gu J, Zhang Z et al (2018) Relation networks for object detection. In: IEEE conference on computer vision and pattern recognition. IEEE, Salt Lake City, UT, pp 3588–3597
Jocher G (2020) Yolov5. Code at https://github.com/ultralytics/yolov5
Kang B, Liu Z, Wang X et al (2019) Few-shot object detection via feature reweighting. In: IEEE international conference on computer vision. IEEE, Seoul, South Korea, pp 8419–8428
Khosla P, Teterwak P, Wang C, et al (2020) Supervised contrastive learning. Preprint at arXiv:2004.11362
Klejch O, Fainberg J, Bell P et al (2019) Speaker adaptive training using model agnostic meta-learning. In: IEEE automatic speech recognition and understanding workshop. IEEE, Singapore, pp 881–888
Li C, Jiang H, Weng K, et al (2022) Yolov6: a single-stage object detection framework for industrial applications. Preprint at arXiv:2209.02976
Liu H, Socher R, Xiong C (2019) Taming MAML: efficient unbiased meta-reinforcement learning. In: Chaudhuri K, Salakhutdinov R (eds) Proceedings of machine learning research, vol 97. PMLR, Long Beach, CA
Liu W, Anguelov D, Erhan D et al (2016) SSD: Single shot multibox detector. In: Leibe B, Matas J, Sebe N et al (eds) Lecture notes in computer science, vol 9905. Springer, Amsterdam, pp 21–37
Munkhdalai T, Yu H (2017) Meta networks. Preprint at arXiv:1703.00837
Nichol A, Achiam J, Schulman J (2018) On first-order meta-learning algorithms. Preprint at arXiv:1803.02999
Redmon J, Farhadi A (2017) Yolo9000: Better, faster, stronger. In: IEEE conference on computer vision and pattern recognition. IEEE, Honolulu, HI, pp 6517–6525
Redmon J, Farhadi A (2018) Yolov3: an incremental improvement. Preprint at arXiv:1804.02767
Redmon J, Divvala S, Girshick R et al (2016) You only look once: unified, real-time object detection. In: IEEE conference on computer vision and pattern recognition. IEEE, Seattle, WA, pp 779–788
Ren S, He K, Girshick R et al (2017) Faster R-CNN: towards real-time object detection with region proposal networks. IEEE Trans Pattern Anal Mach Intell 39:1137–1149
Ruwurm M, Wang S, Krner M et al (2020) Meta-learning for few-shot land cover classification. In: IEEE computer society conference on computer vision and pattern recognition workshops. IEEE, Seattle, WA, pp 788–796
Santoro A, Bartunov S, Botvinick M, et al (2016) One-shot learning with memory-augmented neural networks. Preprint at arXiv:1605.06065
Schweiker K (1993) Fractal detection algorithm for a ladar sensor. In: Proceedings of SPIE—The International Society for Optical Engineering 1960
Shell J, Swersky K, Zemel R (2017) Prototypical networks for few-shot learning. In: Guyon I, Luxburg U, Bengio S et al (eds) Advances in neural information processing systems, vol 30. Curran Associates Inc, Long Beach
Shi B, Liang J, Di L et al (2019) Fabric defect detection via low-rank decomposition with gradient information. IEEE Access 7:130,423-130,437
Sun B, Li B, Cai S et al (2021) FSCE: few-shot object detection via contrastive proposal encoding. In: IEEE conference on computer vision and pattern recognition. IEEE, ELECTR NETWORK, pp 7348–7358
Vanschoren J (2018) Meta-learning: a survey. Preprint at arXiv:1810.03548
Vinyals O, Blundell C, Lillicrap T, et al (2016) Matching networks for one shot learning. Preprint at arXiv:1606.04080
Wang C, Bochkovskiy A, Liao H (2022) Yolov7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. Preprint at arXiv:2207.02696
Wang X, Huang T, Darrell T, et al (2020) Frustratingly simple few-shot object detection. Preprint at arXiv:2003.06957v1
Wang Y, Ramanan D, Hebert M (2019) Meta-learning to detect rare objects. In: IEEE international conference on computer vision. IEEE, Seoul, South Korea, pp 9924–9933
Wu J, Liu S, Huang D, et al (2020) Multi-scale positive sample refinement for few-shot object detection. Preprint at arXiv:2007.09384
Xiao Y, Marlet R (2020) Few-shot object detection and viewpoint estimation for objects in the wild. Preprint at arXiv:2007.12107v1
Yan X, Chen Z, Xu A et al (2019) Meta R-CNN: towards general solver for instance-level low-shot learning. In: IEEE international conference on computer vision. IEEE, Seoul, South Korea, pp 9576–9585
Zhao Y, Yao YSD, et al (2020) Survey on deep learning object detection. J Image Graph 25:5–30
Funding
This work was supported by National Key R &D Program of China (2019YFB1705002), National Natural Science Foundation of China (51634002), LiaoNing Revitalization Talents Program (XLYC2002041).
Author information
Authors and Affiliations
Contributions
HH, XL contributed to Conceptualization, Formal analysis, investigation, Writing—review and editing; HH, XL, CY contributed to Methodology; HH contributed to Writing—original draft preparation; XL contributed to Funding acquisition, Resources and Supervision.
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that there is no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix A The loss curve of GSS and FLGS
Appendix A The loss curve of GSS and FLGS
In meta-learning, meta-training is part of the initial parameter gradient optimization, and meta-testing is part of regular training. Therefore, the epoch in Figs. 12, 13 and 14 contains data from these two stages of data at 0 to 10. It should be noted that there is a period in the curve that looks like loss (epoch from 0 to 10) is rising. It is not an error but a jump. This jump was caused by converting meta-training to meta-testing, which is normal. Observing Figs. 12, 13 and 14, it can be seen that during the meta-training and meta-testing stages, the change trends of the train and val loss curves are consistent, all showing a downward trend. This shows that the training data is effective. The network model training process is normal. There are no problems such as overfitting, underfitting, gradient explosion, and gradient disappearance. The network model is constantly adjusting weights and optimization.
Observing Figs. 12 and 13, it can be found that as the Line increases, the downward trend of the val loss curve becomes more obvious and more convergent. At the same time, GSS plays a full role when Line is greater than 1, and FLGS plays a full role when Line is greater than 2, which indirectly indicates that the multi-line gradient detection strategy is effective. The loss change in the val part of Fig. 14 has some oscillations. This is due to the small amount of data, which is an acceptable normal phenomenon. Their overall downward trend is not problematic.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Huang, H., Luo, X. & Yang, C. Industrial few-shot fractal object detection. Neural Comput & Applic 35, 21055–21069 (2023). https://doi.org/10.1007/s00521-023-08889-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00521-023-08889-z