Siri Whu dataset: Example of the aitlas toolbox in for multi class image classification

This notebook shows a sample implementation of a multi class image classification using the aitlas toolbox and the Siri Whu dataset.

from aitlas.datasets import SiriWhuDataset
from aitlas.models import ResNet50
from aitlas.transforms import ResizeCenterCropFlipHVToTensor, ResizeCenterCropToTensor
from aitlas.utils import image_loader

Load the dataset

dataset_config = {
    "data_dir": "./data/SIRI-WHU",
    "csv_file": "./data/SIRI-WHU/train.csv"
}
dataset = SiriWhuDataset(dataset_config)

Show images from the dataset

fig1 = dataset.show_image(1000)
fig2 = dataset.show_image(80)
fig3 = dataset.show_batch(15)

Inspect the data

dataset.show_samples()

dataset.data_distribution_table()

fig = dataset.data_distribution_barchart()

Load train and test splits

train_dataset_config = {
    "batch_size": 16,
    "shuffle": True,
    "num_workers": 4,
    "data_dir": "./data/SIRI-WHU",
    "csv_file": "./data/SIRI-WHU/train.csv"
}

train_dataset = SiriWhuDataset(train_dataset_config)
train_dataset.transform = ResizeCenterCropFlipHVToTensor()

test_dataset_config = {
    "batch_size": 4,
    "shuffle": False,
    "num_workers": 4,
    "data_dir": "./data/SIRI-WHU",
    "csv_file": "./data/SIRI-WHU/test.csv",
    "transforms": ["aitlas.transforms.ResizeCenterCropToTensor"]
}

test_dataset = SiriWhuDataset(test_dataset_config)
len(train_dataset), len(test_dataset)

Setup and create the model for training

epochs = 10
model_directory = "./experiments/SIRI-WHU"
model_config = {
    "num_classes": 12,
    "learning_rate": 0.0001,
    "pretrained": True,
    "metrics": ["accuracy", "precision", "recall", "f1_score"]
}
model = ResNet50(model_config)
model.prepare()

Training and evaluation

model.train_and_evaluate_model(
    train_dataset=train_dataset,
    epochs=epochs,
    model_directory=model_directory,
    val_dataset=test_dataset,
    run_id='1',
)

Predictions

model_path = "./experiments/SIRI-WHU/checkpoint.pth.tar"
#labels = SiriWhuDataset.labels
labels = ["agriculture", "commercial", "harbor", "idle_land", "industrial", "meadow", "overpass", "park", "pond",
          "residential", "river", "water"]
transform = ResizeCenterCropToTensor()
model.load_model(model_path)

image = image_loader('./data/predict/image1.tif')
fig = model.predict_image(image, labels, transform)

image = image_loader('./data/predict/image2.tif')
fig = model.predict_image(image, labels, transform)

image = image_loader('./data/predict/image3.tif')
fig = model.predict_image(image, labels, transform)

image = image_loader('./data/predict/image4.tif')
fig = model.predict_image(image, labels, transform)