# 1/9/25 Meeting Notes #17
# This Week's Progress 這週進度
## Classification Model #1:
Trained 4 models `訓練了四個模型`:
1. `分主類` classify into dress, upper, lower, and footwear (accuracy: 97.79%)
2. `分類上衣` classify into shirt, t-shirt, hoodie, sweater, polo shirt (accuracy: 90.04%)
3. `分類下裝` classify into pants, shorts, skirt (accuracy: 93.97%)
4. `分類鞋子` classify into flats, heels, shoes, sneakers (accuracy: 88.84%)








- combined 4 models into one 四合一 (accuracy: 90.12%):

## Dataset
- There were multiple duplicates found in dataset for classification model #1
Examples:





- We removed the duplicates. The number of dataset for model 1 decreased from ***43087*** to ***38521***. (about 4.5k duplicates removed)
* 3761 files in 01Dress
* 4017 files in 02Shirt
* 4377 files in 03Tshirt
* ***1492*** files in 04Hoodie
* ***2053*** files in 05Sweater
* **1396** files in 06PoloShirt
* 4899 files in 07Pants
* 3849 files in 08Shorts
* 3022 files in 09Skirt
* 3871 files in 10Flats
* ***265*** files in 11Heels
* ***1703*** files in 12Shoes
* 3816 files in 13Sneakers
* Model2 dataset
* 220 Bohemian
* 1588 Casual
* 419 beach
* 1169 casual
* 1528 Formal
* 1510 Semi-Formal
* 308 buisness casual
* 917 classic
* 285 preppy
* 349 Sporty
* 380 Streetwear
## Backend
- Made the backend for basic account management, upload image (using the model to classify the image), delete item, and edit category (by user). `建立後端,用於基本的帳戶管理、上傳圖片(使用模型對圖片進行分類)、刪除項目,以及編輯分類(由用戶操作)。`
- Tested on a simple frontend (web-based) `在簡單的前端(網頁)進行測試`
# To Do 需做
# Next Weeks's Meeting 下週會議
---
Previous:
Next:
Full Content List [here](https://hackmd.io/@emps-113up/full-list)