fossil-classification

KO | EN

gitlite — search

#python #react #ai #typescript #go #golang #linux #macos #awesome #awesome-list #vue #security #pytorch #open-source

fossil-classification

(★ 9)

to validate whether soft voting could make accuracy increase

fossil-classification 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Overview
Experiment the ensemble technique (soft voting) mentioned in the paper:  
[Fossil Image Identification using Deep Learning
Ensembles of Data Augmented Multiviews](https://arxiv.org/abs/2302.08062)

# Requirement
- pytorch
- pandas
- scikit-learn
- scikit-image
- opencv-python
- tqdm

# Note
I don't completely follow the training hyparameter described in paper, because my mainly purpose is to validate the soft voting.

# Data
- __Download__   
url: https://repository.deep-time.org/detail/1591288605468725250

- __Split the data and transform them into multiviews__  
After unzip the data, type the command below, you will see multiviews which is splitted into `train`, `val`, `test` in the `save_dir`.
```
python dataset.py -data Fusulinid-images-2400\Original_images -save_dir fossil -seed 10
```

# Train
```
python train.py -data fossil -bs 64 -lr 0.01 -end_lr 0.0001 -seed 1000
```

# Evaluate the soft-voting result on the test data
```
python test.py -data fossil -weight weight
```

# Result
- __Training__  
![](https://raw.githubusercontent.com/shyhyawJou/fossil-classification/main/assets/acc.png)

- __Accuracy on test data__

| origin  | soft voting |
  | :-----: | :-----:     |
  | 0.8406 | 0.8844 |  
  
  The soft voting technique really increase the accuracy.

fossil-classification

fossil-classification 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

fossil-classification

fossil-classification 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README