---
tags: 實習日記
---
[TOC]
# 實習日記12/5~12/9

## 篩選照片
1. 特別車牌

2. 歪斜車牌

3. 夜晚車牌

4. 被後車廂擋住

5. 車沒停好

6. 總結
- 歪斜:147張
- 特別車牌:30張
## Label

## train model
1. 訓練數據生成
- https://github.com/Belval/TextRecognitionDataGenerator
- 結果

2. 學習數據轉換
- https://github.com/DaveLogs/TRDG2DTRB

3. 訓練模型
- https://github.com/clovaai/deep-text-recognition-benchmark
- 資料準備好後，還要去選學習模型，EacyOCR用'None-VGG-BiLSTM-CTC'

4. 測試模型

## deep-text-recognition-benchmark
1. Transformation(轉正)
- Transformation (Trans.) normalizes the input text image using the Spatial Transformer Network (STN) to ease downstream stages.
2. Feature extraction(特徵選取)
- Feature extraction (Feat.) maps the input image to a representation that focuses on the attributes relevant for character recognition, while suppressing irrelevant features such as font,color, size, and background.

3. Sequence modeling
- Sequence modeling (Seq.) captures the contextual information within a sequence of characters for the next stage to predict each character more robustly, rather than doing it independently.

4. Prediction (Pred.)
- Prediction (Pred.) estimates the output character sequence from the identified features of an image.
### 參考資料
1. https://www.youtube.com/watch?v=m9RXvCl8wZ8&list=PL3unhkZOa0fY4e3VqQoe4qfMhlmCUW5O7&index=18
2. https://www.youtube.com/watch?v=-j3TbyceShY
3. https://www.youtube.com/watch?v=9_SRXdO9EC4
4. https://zhuanlan.zhihu.com/p/353385040
5. https://zhuanlan.zhihu.com/p/400270506