https://github.com/yukkyo/kaggle-understanding-clouds-69th-solution
69th(Silver) Solution of "Understanding Clouds from Satellite Images | Kaggle" with PyTorch-Lightning
https://github.com/yukkyo/kaggle-understanding-clouds-69th-solution
kaggle python pytorch pytorch-lightning segmentation
Last synced: 3 months ago
JSON representation
69th(Silver) Solution of "Understanding Clouds from Satellite Images | Kaggle" with PyTorch-Lightning
- Host: GitHub
- URL: https://github.com/yukkyo/kaggle-understanding-clouds-69th-solution
- Owner: yukkyo
- Created: 2019-11-08T21:22:54.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2022-11-22T04:37:12.000Z (over 2 years ago)
- Last Synced: 2023-02-26T06:56:38.003Z (about 2 years ago)
- Topics: kaggle, python, pytorch, pytorch-lightning, segmentation
- Language: Python
- Homepage: https://www.kaggle.com/c/understanding_cloud_organization
- Size: 114 KB
- Stars: 19
- Watchers: 1
- Forks: 1
- Open Issues: 7
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Kaggle-Understanding-Clouds-69th-solution
https://www.kaggle.com/c/understanding_cloud_organization/overview
### 0. Environment
- Python 3.6.5
- CUDA 10.1.243
### 1. Preparation
#### 1.1 install Python packages
```bash
$ pip install -r docker/requirements.txt
```
#### 1.2 Download dataset
Use kaggle API command.
```text
input
├── sample_submission.csv
├── test_images
├── train.csv
└── train_images
```
#### 1.3 Split kfolds
```bash
$ cd src
$ python data_process/s01_make_kfold_csv.py --kfold 5
```
### 2. How to train
```bash
$ cd src
$ python train.py --config configs/model063.yaml --kfold 1
$ python train.py --config configs/model063.yaml --kfold 2
$ python train.py --config configs/model063.yaml --kfold 3
$ python train.py --config configs/model063.yaml --kfold 4
$ python train.py --config configs/model063.yaml --kfold 5
$ python train.py --config configs/model064.yaml --kfold 1
$ python train.py --config configs/model064.yaml --kfold 2
$ python train.py --config configs/model064.yaml --kfold 3
$ python train.py --config configs/model064.yaml --kfold 4
$ python train.py --config configs/model064.yaml --kfold 5
```
### 3. How to predict test
```bash
$ cd src
$ python submission.py --kfolds 12345 --config configs/model063.yaml --use-best --save-predicts
$ python submission.py --kfolds 12345 --config configs/model064.yaml --use-best --save-predicts
$ python utils/make_mean_sub.py
$ ls ../output/final
sub_model063064_kfolds12345_top062_minarea16250-13750-10000-5000_bottom042_usebest_6364avg_thres.csv
```