UC Merced dataset: Example of the aitlas toolbox for multi label image classification

This notebook shows a sample implementation of a multi label image classification using the aitlas toolbox using the UC merced multi label dataset.

from aitlas.datasets import UcMercedMultiLabelDataset
from aitlas.models import ResNet50MultiLabel
from aitlas.transforms import ResizeCenterCropFlipHVToTensor, ResizeCenterCropToTensor
from aitlas.utils import image_loader

Load the dataset

dataset_config = {
    "data_dir": "/media/hdd/multi-label/UCMerced_multilabel/images",
    "csv_file": "/media/hdd/multi-label/UCMerced_multilabel/multilabels.txt"
}
dataset = UcMercedMultiLabelDataset(dataset_config)

Show images from the dataset

fig1 = dataset.show_image(1000)
fig2 = dataset.show_image(30)
fig3 = dataset.show_batch(15)

Inspect the data

dataset.show_samples()

	IMAGE\LABEL	airplane	bare-soil	buildings	cars	chaparral	court	dock	field	grass	mobile-home	pavement	sand	sea	ship	tanks	trees	water
0	agricultural00	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	1	0
1	agricultural01	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
2	agricultural02	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
3	agricultural03	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
4	agricultural04	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0
5	agricultural05	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
6	agricultural06	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
7	agricultural07	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
8	agricultural08	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
9	agricultural09	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
10	agricultural10	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
11	agricultural11	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
12	agricultural12	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
13	agricultural13	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
14	agricultural14	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
15	agricultural15	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
16	agricultural16	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
17	agricultural17	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
18	agricultural18	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0
19	agricultural19	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0

dataset.data_distribution_table()

	Label	Count
1	airplane	100
2	bare-soil	718
3	buildings	691
4	cars	886
5	chaparral	115
6	court	105
7	dock	100
8	field	103
9	grass	975
10	mobile-home	102
11	pavement	1300
12	sand	294
13	sea	100
14	ship	102
15	tanks	100
16	trees	1009
17	water	203

fig = dataset.data_distribution_barchart()

Load train and test splits

train_dataset_config = {
    "batch_size": 16,
    "shuffle": True,
    "num_workers": 4,
    "data_dir": "/media/hdd/multi-label/UCMerced_multilabel/images",
    "csv_file": "/media/hdd/multi-label/UCMerced_multilabel/train.csv"
}

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

test_dataset_config = {
    "batch_size": 4,
    "shuffle": False,
    "num_workers": 4,
    "data_dir": "/media/hdd/multi-label/UCMerced_multilabel/images",
    "csv_file": "/media/hdd/multi-label/UCMerced_multilabel/test.csv",
    "transforms": ["aitlas.transforms.ResizeCenterCropToTensor"]
}

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

(1683, 417)

Setup and create the model for training

epochs = 10
model_directory = "/media/hdd/multi-label/UCMerced_multilabel/experiments/"
model_config = {
    "num_classes": 17,
    "learning_rate": 0.0001,
    "pretrained": True,
    "threshold": 0.5,
    "metrics": ["accuracy", "precision", "recall", "f1_score"]
}
model = ResNet50MultiLabel(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 = "/media/ssd/aitlas_new/examples/experiment/uc_merced_test/checkpoint.pth.tar"
#labels = UcMercedMultiLabelDataset.labels
labels = ["airplane", "bare-soil", "buildings", "cars", "chaparral", "court", "dock", "field", "grass",
                "mobile-home", "pavement", "sand", "sea", "ship", "tanks", "trees", "water"]
transform = ResizeCenterCropToTensor()
model.load_model(model_path)

image = image_loader('/media/ssd/uc_merced_multilabel/predict/buildings98.tif')
fig = model.predict_image(image, labels, transform)

image = image_loader('/media/ssd/uc_merced_multilabel/predict/golfcourse04.tif')
fig = model.predict_image(image, labels, transform)

image = image_loader('/media/ssd/uc_merced_multilabel/predict/airplane18.tif')
fig = model.predict_image(image, labels, transform)

image = image_loader('/media/ssd/uc_merced_multilabel/predict/parkinglot13.tif')
fig = model.predict_image(image, labels, transform)

2021-11-24 21:50:52,611 INFO Loading checkpoint /media/ssd/aitlas_new/examples/experiment/uc_merced_test/checkpoint.pth.tar
2021-11-24 21:50:52,796 INFO Loaded checkpoint /media/ssd/aitlas_new/examples/experiment/uc_merced_test/checkpoint.pth.tar at epoch 51