Additional file 1 of Deep learning image segmentation for the reliable porosity measurement of high-capacity Ni-based oxide cathode secondary particles
Additional file 1: Fig. S1. Effect of FFT and Gaussian filters for preparing input images. a Cross-sectional backscatter electron (BSE) SEM image of an NCM secondary particle as acquired without filtering. The vertical stripe-type contrast artifacts are formed due to the curtaining effect. b FFT filtered BSE SEM image to remove vertical stripe artifacts. c Prepared input image after removing statistical noise background by Gaussian filter for the FFT filtered image. Fig. S2. Application of the trained model to a different NCA secondary particle for detection and segmentation performance for irregular shape pores. This extra data supports that
Resource information
- Authors
- Lee, Hee-Beom, Jung, Min-Hyoung, Kim, Young-Hoon, Park, Eun-Byeol, Jang, Woo-Sung, Kim, Seon-Je, Choi, Ki-ju, Park, Ji-young, Hwang, Kee-bum, Shim, Jae-Hyun, Yoon, Songhun, Kim, Young-Min
- Research area
- Artificial intelligence
- Research topics
- Artificial Intelligence and Image Processing, FOS: Computer and information sciences
- Publication date
- 2023
- DOI
- 10.6084/m9.figshare.24603582
- Upstream last modified
- Feb 27, 2026
- License
- Creative Commons Attribution 4.0 International
- Last record update
- Nov 5, 2026
About this resource
Additional file 1: Fig. S1. Effect of FFT and Gaussian filters for preparing input images. a Cross-sectional backscatter electron (BSE) SEM image of an NCM secondary particle as acquired without filtering. The vertical stripe-type contrast artifacts are formed due to the curtaining effect. b FFT filtered BSE SEM image to remove vertical stripe artifacts. c Prepared input image after removing statistical noise background by Gaussian filter for the FFT filtered image. Fig. S2. Application of the trained model to a different NCA secondary particle for detection and segmentation performance for irregular shape pores. This extra data supports that the accuracy of >97% for feature segmentation is well maintained, thus suggesting the model’s generalizability. Fig. S3. Comparison of pore segmentation performance between human-driven process with preprocessed image dataset and CNN-assisted process with unprocessed image dataset. Without preprocessing, false-positive detection in feature segmentation was notably increased for using the CNN-assisted process. This result indicates the importance of preprocessing of input images for deep learning-assisted pore segmentation.