We use cGAN to fillin the synthetic colors on gray images of border/vein. And evaluated the reconstruction accuracy by leaf types classification using Alexnet CNN Protopytpe of generate fake image from hand-drawn vein has also been proposed. See in detail: Report, PPT
－ Step 1. Vein detection (Pre-processing)
－ Step 2. GAN (GAN reconstruction using pre-processed image)
－ Step 3. CNN (Classification using Alexnet)
Addition: Interactive Interface, allows to draw images in real time, and get GAN reconstruction, and classification using CNN.
Code adapted from cycleGAN https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix
Task Distribution
| Group Member | Contributions |
|---|---|
| Rachel Mills | Literature review, dataset assembly |
| Raj Shah | GAN model study, GAN result analysis, report compilation |
| Xiaoyang Li | Image preprocessing, GAN code review, GAN implementation |
| Gaurav Roy | CNN model study, CNN result analysis, report compilation |
| Aditi Singh | GAN code review, CNN implementation, Interactive Interface building |