Structured Filter Pruning applied to Mask R-CNN: A path to efficient image segmentation
DOI:
https://doi.org/10.26034/lu.akwi.2024.4673Keywords:
Pruning, Filter Pruning, Mask R-CNN, Image Segmentation, Two-stage detectorsAbstract
This study investigates the optimization of two-stage object recognition systems, integral to a variety of applications, through advanced machine learning techniques. While such systems achieve high accuracy, they often suffer from over-parametrization and excessive computational demands. Addressing this, a pruning method tailored for complex architectures like Mask R-CNN was developed, aimed at reducing model complexity and storage without sacrificing accuracy. Employing a Global Kernel Level Filter Pruning strategy, identified by the L1-Norm, the approach strategically eliminated non-essential parameters post-training. This method maintains high recognition accuracy; the experiments demonstrated that accuracy remained robust up to 40\% pruning, achieving a compression ratioof 1.25 and an IoU@0.5 of 0.72. The findings offer valuable insights for the field, presenting a nuanced approach to enhancing Neural Network efficacy while maintaining performance, thereby contributing to the ongoing dialogue in AI optimization.References
Abbasi-Asl, R. and B. Yu 2017. “Structural Compression of Convolutional Neural Networks Based on Greedy Filter Pruning”. In: arXiv preprint arXiv:1705.07356.
Aguiar Salvi, A. de and R. C. Barros 2021. “Model Compression in Object Detection”. In: 2021 International Joint Conference on Neural Networks (IJCNN). IEEE, pp. 1–8.
Alpaydin, E. 2016. Machine learning, The new AI. Cambridge, MA: MIT Press.
Anwar, S. et al. 2017. “Structured Pruning of Deep Convolutional Neural Networks”. In: ACM Journal on Emerging Technologies in Computing Systems 13.3, pp. 1–18.
Basili, V. R. 2002. “The role of experimentation in software engineering: past, current, and future”. In: Proceedings of IEEE 18th International Conference on Software Engineering. 1, pp. 1–8.
Blalock, D. et al. 2020. “What is the State of Neural Network Pruning?” In: Proceedings of Machine Learning and Systems 2020 1.1, pp. 1–18.
Canziani, A. et al. 2016. An Analysis of Deep Neural Network Models for Practical Applications. arXiv.
Chen, L. et al. 2021. “Knowledge from the original network: restore a better pruned network with knowledge distillation”. In: Complex Intelligent Systems 8.1, pp. 1–10.
Cosman, P. C. et al. 1993. “Using vector quantization for image processing”. In: Proceedings of the IEEE 81.9, pp. 1326–1341.
Facebook 2020. From Research to Production. https://pytorch.org/. Retrieved: 05.05.2020.
Feng, Di et al. 2021. “A Review and Comparative Study on Probabilistic Object Detection in Autonomous Driving”. In: IEEE Transactions on Intelligent Transportation Systems 1, pp. 1–20.
Feng, M. et al. 2013. Complementarity formulations of l0-norm optimization problems. Industrial Engineering and Management Sciences. Technical Report.
Gholami, A. et al. 2015. “A Survey of Quantization Methods for Efficient Neural Network Inference”. In: Proceedings of the IEEE International Conference on Computer Vision. Santiago, pp. 1440–1448.
H. Zhou, et. al 2016. “Less Is More: Towards Compact CNNs”. In: ECCV. Springer, pp. 662–677.
Han, S. et. al 2016. “Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding”. In: International Conference on Learning Representations. San Juan, pp. 1–14.
Hassibi, B. and D. Stork 1992. “Second Order Derivatives for Network Pruning: Optimal Brain Surgeon”. In: Neural Information Processing Systems 1992, pp. 164–171.
Hu, H. et al. 2016. “Network Trimming: A Data-Driven Neuron Pruning Approach Towards Efficient Deep Architectures”. In: arXiv preprint arXiv:1607.03250.
Huang, J. et al. 2017. “Speed/Accuracy Trade-Offs for Modern Convolutional Object Detectors”. In: IEEE Conference on Computer Vision and Pattern Recognition 2017, pp. 1–21.
Kumar, A. et al. 2021. “Pruning filters with L1-norm and capped L1-norm for CNN compression”. In: Appl Intell 51, pp. 1152–1160. doi: 10.1007/s10489-020-01894-y.
LeCun, Y. et al. 1989. “Optimal Brain Damage”. In: Advances in Neural Information Processing Systems. Ed. by D. Touretzky. Morgan-Kaufmann, pp. 1–8.
Li, H. et al. 2017. “Pruning Filters for Efficient ConvNets”. In: International Conference on Learning Representations. Toulon, pp. 1–13.
Lin, T.-Y. et al. 2014. “Microsoft COCO: Common Objects in Context”. In: European Conference on Computer Vision. Springer, pp. 740–755.
Liu, Z. et al. 2017. “Learning Efficient Convolutional Networks Through Network Slimming”. In: ICCV. IEEE, pp. 2755–2763.
Molchanov, P. et al. 2017. “Pruning Convolutional Neural Networks for Resource Efficient Inference”. In: ICLR.
Peffers, K. et al. 2007. “A Design Science Research Methodology for Information Systems Research”. In: Journal of Management Information Systems 24.3, pp. 45–77.
Rastegari, M. et al. 2016. “XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks”. In: ECCV. Springer, pp. 525-542.
Shinde, P. P. and S. Shah 2018. “A Review of Machine Learning and Deep Learning Applications”. In: 2018 Fourth International Conference on Computing Communication Control and Automation (ICCUBEA). IEEE, pp. 1–6.
Singh, P. et al. 2019. “Stability Based Filter Pruning for Accelerating Deep CNNs”. In: Winter Conference on Applications of Computer Vision 2019. IEEE. Waikoloa Village, pp. 1–9.
Tzelepis, G. et al. 2019. Deep Neural Network Compression for Image Classification and Object Detection. n.d.: n.p..
Wen, W. et al. 2016. “Learning Structured Sparsity in Deep Neural Networks”. In: NIPS, pp. 2074–2082.
Zou, Z. et al. 2019. Object Detection in 20 Years: A Survey. n.d.: n.p.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 Yannik Frühwirth
This work is licensed under a Creative Commons Attribution 4.0 International License.
