MercyMarkus

Kagoro People

A simple Tribute Page to my Unknown Tribe.

A Pen by MercyMarkus on CodePen.

License.

MercyMarkus

# This ensures that any edits to libraries you make are reloaded here automatically,
# and also that any charts or images displayed are shown in this notebook.
%reload_ext autoreload
%autoreload 2
%matplotlib inline

MercyMarkus

# Import libraries
from fastai import *
from fastai.vision import *
from fastai.callbacks import CSVLogger, SaveModelCallback
import seaborn as sns
import warnings
warnings.filterwarnings('ignore')

MercyMarkus

# Get the data from dropbox link
!wget https://www.dropbox.com/s/6kltw0kqynlijxv/widsdatathon2019.zip
  
# The downloaded competition data is zipped, let us unzip it
!unzip widsdatathon2019.zip

# The training and testing data have already been seperated, Unzip them as well
!unzip train_images.zip
!unzip leaderboard_holdout_data.zip
!unzip leaderboard_test_data.zip

MercyMarkus

# Overview of the labels of the training data; 
df = pd.read_csv('data/traininglabels.csv')
df.head()

MercyMarkus

src = (ImageList.from_df(df, path, folder='train_images')
      .random_split_by_pct(0.2, seed=14)
      .label_from_df('has_oilpalm')
      .add_test(combined_test))
      
data = (src.transform(get_transforms(flip_vert=True), size=164)
           .databunch()
           .normalize(imagenet_stats))

MercyMarkus

test_imgs = [i for i in test.iterdir()]
hold_imgs = [i for i in lb_test.iterdir()]
combined_test = test_imgs + hold_imgs
len(combined_test)

MercyMarkus

sns.countplot(df.has_oilpalm)

MercyMarkus

data.show_batch(2)

MercyMarkus

learn = create_cnn(data, models.resnet50, metrics=[accuracy, error_rate],
                  callback_fns=[ShowGraph, SaveModelCallback])
# View model architecture
learn.model()