Defects in Li-ion batteries on a microstructural level such as foreign inclusions, cracks or delaminations can have implications on a battery´s performance, safety as well as ageing behavior [1]. The search for these kinds of defects and anomalies in assembled cells requires high resolution images, resulting in hundreds or thousands of images per cell to be evaluated. Earlier work already showed that deep learning methods like convolutional neural networks (CNN) can successfully be applied to classify micrographs to detect defects in Li-ion battery microstructures [2].
In parallel, significant advances have been made in the field of computer vision by adopting Transformers [3] (an architecture originally designed for natural language processing) for image classification tasks. Approaches like Vision Transformers (ViT) [4] and Swin Transformers [5] outperformed their state-of-the art CNN counterparts. Design principles from these architectures as well as other improvements found their way back into pure CNN architectures. In [6], Liu et al. studied the influence of these improvements when collectively applied and achieved a new state-of-the art for computer vision tasks with their new family of CNN models – namely ConvNeXt. The proposed changes   include model training (optimization, regularization) as well as model design in terms of activation functions, normalization, depth-wise separable convolutions, kernel sizes, inverted bottlenecks, among others.
In this work, we investigate the influence of adopting the ConvNeXt design principles as proposed in [6] on the task of defect detection in micrographs of Li-ion batteries, looking at training behavior as well as classification accuracy and their perspectives for an efficient and deep learning-based microscopical battery evaluation.

References
[1]    G. Qian et al., “The role of structural defects in commercial lithium-ion batteries,” Cell Reports Physical Science, vol. 2, no. 9, p. 100554, 2021, doi: 10.1016/j.xcrp.2021.100554.
[2]    O. Badmos, A. Kopp, T. Bernthaler, and G. Schneider, “Image-based defect detection in lithium-ion battery electrode using convolutional neural networks,” J Intell Manuf, vol. 31, no. 4, pp. 885–897, 2020, doi: 10.1007/s10845-019-01484-x.
[3]    A. Vaswani et al., “Attention Is All You Need,” Jun. 2017. [Online]. Available: https://arxiv.org/pdf/1706.03762
[4]    A. Dosovitskiy et al., “An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale,” Oct. 2020. [Online]. Available: https://arxiv.org/pdf/2010.11929
[5]    Z. Liu et al., “Swin Transformer: Hierarchical Vision Transformer using Shifted Windows,” Mar. 2021. [Online]. Available: https://arxiv.org/pdf/2103.14030
[6]    Z. Liu, H. Mao, C.-Y. Wu, C. Feichtenhofer, T. Darrell, and S. Xie, “A ConvNet for the 2020s,” Jan. 2022. [Online]. Available: https://arxiv.org/pdf/2201.03545