Intelligent Information Processing Lab – Seminar Insights
The Intelligent Information Processing Lab seminar delved into advanced techniques such as deep learning approaches, geometric transformations, color space transformations, and adversarial training. With a focus on plant image data augmentation, the lab showcased innovative methodologies such as GAN data augmentation and neural style transfer. Groundbreaking research results were presented, including improvements in plant disease classification using generative adversarial networks and effective data augmentation methods for plant disease diagnosis.
Download Presentation
Please find below an Image/Link to download the presentation.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.
You are allowed to download the files provided on this website for personal or commercial use, subject to the condition that they are used lawfully. All files are the property of their respective owners.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.
E N D
Presentation Transcript
Lab Seminar Intelligent Information Processing Lab 2023. 03. 29
IIP Lab Plant Image Data Augmentation Basic Image Manipulations Deep Learning Approaches Geometric Transformations Color Space Transformations Adversarial Training GAN Data Augmentation Kernel Filters Neural Style Transfer Random Erasing Mixing Images 2
IIP Lab Plant (2020) WGAN-GP-LSR(Lebel Smoothing Regularization) + CNN : 14 38 / Pure CNN : 60% CNN + Basic Image Manipulation + WGAN-GP : 82% CNN + Basic Image Manipulation + WGAN-GP + LSP : 84% Rotation + Brightness+ Contrast (2020) Leafgan + ResNet : 1 ( ) 4 / Pure ResNet : 97.2% ResNet + Basic Image Manipulation + CycleGAN : 97.7%% ResNet + Basic Image Manipulation + LeafGAN : 97.9% Rotation + Horizontal flip + Vertical flip 3 1. 2. L. Bi and G. Hu, "Improving image-based plant disease classification with generative adversarial network under limited training set", Frontiers Plant Sci., vol. 11, pp. 1945, Dec. 2020. Cap, Quan Huu, et al. "Leafgan: An effective data augmentation method for practical plant disease diagnosis." IEEE Transactions on Automation Science and Engineering 19.2 (2020): 1258- 1267.
IIP Lab Plant (2021) C-GAN + DenseNet : 1 ( ) 10 / DenseNet C-GAN + DenseNet 5 98% 99% 7 94% 98% 10 93% 97% Problem WGAN-GP-LSP : Leafgan + C-GAN : ( ) DL Data augmentation color distribution x 4 3. A. Abbas, S. Jain, M. Gour and S. Vankudothu, "Tomato plant disease detection using transfer learning with C-GAN synthetic images", Comput. Electron. Agricult., vol. 187, Aug. 2021.
Plant 5 26 / Model f1 score accuracy f1 score accuracy Training time VGGNet16 97.2% 97.4% 97.2% 98.4% 1 day 14:09:16 ResNet50 95.3% 94.7% 97.3% 98.6% 1 day 14:26:13 DenseNet161 98.4% 98.3% 98.8% 99.16% 1 day 18:33:52 EfficientNet 97.5% 97.6% 97% 98.5% 1 day 14:02:07 ViT 98.3% 98% 98.8% 99.12% 1 day 23:44:38 DeiT(none) 97.2% 97% DeiT(hard) 97.4% 97% 5
IIP Lab Plant Horizontal, Vertical , Rotation, Resize and crop Stratified kfold (fold = 5) Color jitter, Gaussian blur, Rotation, Resize and crop C-GAN C-GAN + ( ) C-GAN + ( ) 6
IIP Lab Plant Contribution data augmentation , k-fold cross validation GAN+CNN Transformer DL GAN data augmentation Kiwi (ACM_RACS 2023) k-fold cross validation Model F1 score(avg) Acc(avg) Training time(total) VGG 97.1% 97.7% 22:45:20 ResNet 96.46% 97.56% 22:45:45 7 DenseNet 97.18% 97.72% 22:57:43
Thank you SaeBom Lee Department of Computer Engineering, Gachon University | Researcher Tel. +82-31-750-8822 Mobile. +82-10-6641-9390 E-mail. dltoqha@gachon.ac.kr
