Skip to content

Simple_FastAI_exercise

Author: https://www.kaggle.com/kenseitrg

From: https://www.kaggle.com/kenseitrg/simple-fastai-exercise

License: Apache 2.0

Score: 1.0000

In [1]:

import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)

In [2]:

from pathlib import Path
from fastai import *
from fastai.vision import *
import torch

In [3]:

data_folder = Path("../input")
#data_folder.ls()

In [4]:

train_df = pd.read_csv("../input/train.csv")
test_df = pd.read_csv("../input/sample_submission.csv")

In [5]:

test_img = ImageList.from_df(test_df, path=data_folder/'test', folder='test')
trfm = get_transforms(do_flip=True, flip_vert=True, max_rotate=10.0, max_zoom=1.1, max_lighting=0.2, max_warp=0.2, p_affine=0.75, p_lighting=0.75)
train_img = (ImageList.from_df(train_df, path=data_folder/'train', folder='train')
        .split_by_rand_pct(0.01)
        .label_from_df()
        .add_test(test_img)
        .transform(trfm, size=128)
        .databunch(path='.', bs=64, device= torch.device('cuda:0'))
        .normalize(imagenet_stats)
       )

In [6]:

#train_img.show_batch(rows=3, figsize=(7,6))

In [7]:

learn = cnn_learner(train_img, models.densenet161, metrics=[error_rate, accuracy])


Downloading: "https://download.pytorch.org/models/densenet161-8d451a50.pth" to /tmp/.torch/models/densenet161-8d451a50.pth
115730790it [00:06, 17660476.43it/s]

In [8]:

#learn.lr_find()
#learn.recorder.plot()

In [9]:

lr = 3e-02
learn.fit_one_cycle(5, slice(lr))

Total time: 06:22

epoch train_loss valid_loss error_rate accuracy time
0 0.064186 0.090407 0.017143 0.982857 01:25
1 0.039902 0.001073 0.000000 1.000000 01:15
2 0.027812 0.003891 0.000000 1.000000 01:13
3 0.012305 0.000548 0.000000 1.000000 01:14
4 0.002986 0.001019 0.000000 1.000000 01:14

In [10]:

#learn.unfreeze()
#learn.lr_find()
#learn.recorder.plot()

In [11]:

#learn.fit_one_cycle(1, slice(1e-06))

In [12]:

#interp = ClassificationInterpretation.from_learner(learn)
#interp.plot_top_losses(9, figsize=(7,6))

In [13]:

preds,_ = learn.get_preds(ds_type=DatasetType.Test)

In [14]:

test_df.has_cactus = preds.numpy()[:, 0]

In [15]:

test_df.to_csv('submission.csv', index=False)
回到顶部