# 05/01/25 Meeting Notes #30
# This Week's Progress 這週進度
## Garment Classification
- Due to insufficient templates for footwear in 2D and 3D visualization, we have decided to remove footwear entirely from our project.
- Updated accuracies (before and after thresholding), confidence threshold
- Accuracies:
| Class | Before | Threshold | After |
| -------- | -------- | -------- | ---|
| Main Category | 97.97% | 0.90 | 99.34|
| Top Subcategory | 88.99% | 0.53| 91.54%|
| Bottom Subcategory| 92.34% | 0.59 | 94.28% |
| Overall | 89.81% | 0.59 | 92.20 |
-- Before removing footwear, the accuracy for overall after threshold was 92.29%%
- Graphs:
| Class | curve | before | after |
| ----- | -------- | -------- | -------- |
| Main Category |  |  |  |
| Top Subcategory |  |  |  |
| Bottom Subcategory|  |  |  |
|Overall | |  |  |

## Style Classification
- Retrained using "Bohemian" instead of "Bohemian and Hippie", "Bohemian and Hippie" is very long, does not look good on the app. Even though bohemian and hippie are two different style, most people cannot tell the difference between the two categories.
- We also tried training with "Bohemian" dataset. As the accuracies between with and without "Bohemian" is low. We've decided to continue using "Bohemian".
- Accuracies
| Bohemian | Before | Threshold | After |
| -------- | -------- | -------- | ---- |
| NO | 72.74% | 0.46 |75.09%|
| YES | 74.86% | 0.45 |76.76%|

- Graphs:
| Bohemian | Loss Graph | Before Threshold | After Threshold |
| -------- | -------- | -------- | ----- |
| NO | |  | |
| YES || ||

## Frontend
- Connected some backend api: Sign Up, OTP, Sign In, upload images, call images)
- Remaining functions to connect: show 2d visualization of vton, show 3d styles, log out.
- Remaining things to do: prepare girl mannequin, add ways for user to sign in using google, delete garment, edit garment, reposition everything to align with Elvio's phone.

## Paper
- [HERE](https://docs.google.com/document/d/1WPQakHBWTsFGSDrFCPxyDMCFHCaDIV9G/edit?usp=sharing&ouid=109251808530722079781&rtpof=true&sd=true)
- The current paper no longer includes footwear.
- Done: Input process after user upload(section 4.1, 5.1), Garment classification(Section 4.2, 5.2), 2D visualization (Section 4.3, 5.3), Style classification (Section 4.4, 5.4), 3D Visualization (Section 4.5, 5.5), Discussion (not including the user interface), Appendix A (Loss Graph), Appendix B (Confusion Matrix)
- Not done: Related works(section 3), future work(section 8), references(section 9), computational environment specification (section 4.5.6), user interface (app)(Section 4.6), Qualitative performance analysis (Section 5.4.3, needs to rewrite again), conclusion (section 7)
## Ppt
- [HERE](https://www.canva.com/design/DAGl2z1t9iI/wSbc7uGtN9NXVjjzs1Ekkw/edit?utm_content=DAGl2z1t9iI&utm_campaign=designshare&utm_medium=link2&utm_source=sharebutton )
- Still in progress
## Poster
- The poster currently looks like this.
- Still Discussing what to put on the poster.
- Might change the positions a bit.

# To Do 需做
- Decrease paper length, too long.

# Next Weeks's Meeting 下週會議

---
Previous:
Next:
Full Content List [here](https://hackmd.io/@emps-113up/full-list)