This project was conducted as a part of the 2018 Fall Creative Integrated Design Course at Seoul National University, which aimed to provide opportunities to participate in university-industry projects.
I was involved in developing a plant disease detection system & web application as a team of three. Our team collaborated with LG Farm Hannong, the biggest agricultural company in South Korea, under the supervision of Prof. Yeong-Gil Shin, SNU. My role was implementing and testing deep learning models, and visualizing the model interpretation, and implementing the backend server.
The plant disease detection was tackled by solving the image classification problem, which categorizes the pathological degree of each disease (e.g. unaffected/early/terminal marssonia blotch). As the dataset provided from Farm Hannong was limited, we also used a public plant disease dataset from PlantVillage  to facilitate transfer learning . ResNet  with 34 layers showed the best performance (96.89 %) on validation data in the experiment.
To improve users' confidence in our system, we also implemented a feature visualization. We utilized Guided GRAD-CAM  techniques to highlight which parts of the plant image are attributed to the specific classification results.
Finally, our team implemented the web application providing plant disease detection & visualization, and relevant product recommendation.
 Hughes, D., & Salathé, M. (2015). An open access repository of images on plant health to enable the development of mobile disease diagnostics. arXiv preprint arXiv:1511.08060.
 Yosinski, J., Clune, J., Bengio, Y., & Lipson, H. (2014). How transferable are features in deep neural networks?. In Advances in neural information processing systems (pp. 3320-3328).
 He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 770-778).
 Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., & Batra, D. (2017). Grad-cam: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE international conference on computer vision (pp. 618-626).