AI_Binbot/training_convnext.py at main · Mariposas-Inteligentes/AI_Binbot · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
def main():
    import os
    import sys
    import inspect

    currentdir = os.path.dirname(
        os.path.abspath(inspect.getfile(inspect.currentframe()))
    )
    parentdir = os.path.dirname(currentdir)
    sys.path.insert(0, parentdir)

    import pandas as pd
    import torch
    from pytorch_lightning.loggers import WandbLogger

    from helper_functions import count_classes

    from models.image_classification_lightning_module import (
        ImageClassificationLightningModule,
    )

    from pytorch_lightning import Trainer

    from pytorch_lightning.callbacks import ModelCheckpoint

    from pytorch_lightning.callbacks import EarlyStopping

    from data.data_module import TrashDataModule, Sampling

    from torchmetrics.classification import (
        MulticlassAccuracy,
        MulticlassConfusionMatrix,
        MulticlassPrecision,
        MulticlassRecall,
    )

    from torchmetrics import MetricCollection
    from models.convnext import ConvNext
    from torch import nn
    import wandb
    import configuration as config

    device = "cuda" if torch.cuda.is_available() else "cpu"
    # ray tracing core
    torch.set_float32_matmul_precision("high")

    class_count = count_classes(config.ROOT_DIR)

    metrics = MetricCollection(
        {
            "Accuracy": MulticlassAccuracy(num_classes=class_count, average="micro"),
            "BalancedAccuracy": MulticlassAccuracy(num_classes=class_count),
            "Precision": MulticlassPrecision(num_classes=class_count),
            "Recall": MulticlassRecall(num_classes=class_count),
        }
    )

    vector_metrics = MetricCollection(
        {
            "Accuracy": MulticlassAccuracy(num_classes=class_count, average=None),
            "Precision": MulticlassPrecision(num_classes=class_count, average=None),
            "Recall": MulticlassRecall(num_classes=class_count, average=None),
            "Confusion Matrix": MulticlassConfusionMatrix(num_classes=class_count),
        }
    )

    train_transform, test_transform = ConvNext.get_transformations()

    trash_data_module = TrashDataModule(
        root_dir=config.ROOT_DIR,
        batch_size=config.BATCH_SIZE,
        test_size=config.TEST_SIZE,
        use_index=config.USE_INDEX,
        indices_dir=config.INDICES_DIR,
        sampling=Sampling.NONE,
        train_transform=train_transform,
        test_transform=test_transform,
    )

    trash_data_module.prepare_data()
    trash_data_module.create_data_loaders()

    metrics_data = []

    convnext = ConvNext(num_classes=class_count, device=device)

    model = ImageClassificationLightningModule(
        model=convnext,
        loss_fn=nn.CrossEntropyLoss(),
        metrics=metrics,
        vectorized_metrics=vector_metrics,
        lr=config.LR,
        scheduler_max_it=config.SCHEDULER_MAX_IT,
    )

    early_stop_callback = EarlyStopping(
        monitor="val/loss",
        patience=config.PATIENCE,
        strict=False,
        verbose=False,
        mode="min",
    )

    checkpoint_callback = ModelCheckpoint(
        monitor="val/loss",
        dirpath=config.CONVNEXT_DIR,
        filename=config.CONVNEXT_FILENAME,
        save_top_k=config.TOP_K_SAVES,
        mode="min",
    )

    id = config.CONVNEXT_FILENAME + "_" + wandb.util.generate_id()
    wandb_logger = WandbLogger(project=config.WANDB_PROJECT, id=id, resume="allow")

    trainer = Trainer(
        logger=wandb_logger,
        callbacks=[early_stop_callback, checkpoint_callback],
        max_epochs=config.EPOCHS,
        log_every_n_steps=1,
    )

    trainer.fit(model, datamodule=trash_data_module)

    # save the metrics per class as well as the confusion matrix to a csv file
    metrics_data.append(trainer.test(model, datamodule=trash_data_module)[0])

    results_per_class_metrics = model.test_vect_metrics_result

    metrics_per_class = pd.DataFrame(
        {
            "Accuracy": results_per_class_metrics["Accuracy"].cpu().numpy(),
            "Precision": results_per_class_metrics["Precision"].cpu().numpy(),
            "Recall": results_per_class_metrics["Recall"].cpu().numpy(),
        },
        index=config.CLASS_NAMES,
    )

    confusion_matrix = pd.DataFrame(
        results_per_class_metrics["Confusion Matrix"].cpu().numpy(),
        index=config.CLASS_NAMES,
        columns=config.CLASS_NAMES,
    )

    metrics_per_class.to_csv(config.CONVNEXT_CSV_PER_CLASS_FILENAME)
    confusion_matrix.to_csv(config.CONVNEXT_CSV_CM_FILENAME)
    wandb.finish()

    pd.DataFrame(metrics_data).to_csv(config.CONVNEXT_CSV_FILENAME, index=False)


if __name__ == "__main__":
    main()