CropWaterPredict/draw_inference.py at main · wantbook-book/CropWaterPredict · 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
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
import models.sapflow_predict as sfp_models
import models.soil_moisture_predict as smp_models
import dataset.sapflow_predict as sfp_datasets
import dataset.soil_moisture_predict as smp_datasets
import torch
from pathlib import Path
import numpy as np
from utils.inference import compare_models, read_labels, draw_infer_graph, cal_r2
from torch.utils.data import DataLoader
import matplotlib.pyplot as plt
from sklearn.metrics import r2_score

def infer_and_draw():
    device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
    batch_size = 32
    num_workers = 12
    src_dir = Path(__file__).resolve().parent
    # -----需要改------
    model1 = smp_models.RgbVgg16Model()
    model1.load_state_dict(torch.load(src_dir/'train/soil_moisture_predict/segment_rgb_vgg16/14r/best_weights.pth', map_location=device))
    # model1 = sfp_models.RgbResNet18TmModel()
    # model1.load_state_dict(torch.load(src_dir/'train/sapflow_predict/rgb_resnet18_tm/best1r/best_weights.pth', map_location=device))
    # model2 = smp_models.RgbVgg16Model()
    # model2.load_state_dict(torch.load(src_dir/'train/soil_moisture_predict/segment_rgb_vgg16/14r/best_weights.pth', map_location=device))
    # model2 = sfp_models.RgbResNet18TmTransformerModel()
    # model2.load_state_dict(torch.load(src_dir/'train/sapflow_predict/rgb_resnet18_tm_transformer/17r/best_weights.pth', map_location=device))
    model2 = model1
    # -----需要改------

    rgb_images_dir = src_dir / 'data' / 'rgb_images'
    segment_images_dir = src_dir / 'data' / 'segment_rgb_images'
    # infrared_image_dir = src_dir / 'data' / 'thermal_data_processed'
    temp_moisture_file_path = src_dir / 'data' / 'series_data' / 'T_moisture_data.csv'
    soil_moisture_file_path = src_dir / 'data' / 'labels' / 'soil_water_content.CSV'
    sapflow_rgb_images_dir = src_dir / 'data' / 'sapflow_predict_data' / 'rgb_images'
    sapflow_dir = src_dir / 'data' / 'sapflow_predict_data' / 'sapflow'

    # -----需要改------
    # dataset = sfp_datasets.RgbTmDataset(
    #     rgb_images_dir=sapflow_rgb_images_dir,
    #     tm_file_path=temp_moisture_file_path,
    #     sapflow_dir=sapflow_dir,
    #     transform=model1.get_image_transform(is_training=False)
    # )
    dataset = smp_datasets.rgb_tm_dataset(
        # rgb_images_dir=rgb_images_dir,
        rgb_images_dir=segment_images_dir,
        temp_moisture_filepath=temp_moisture_file_path,
        soil_moisture_filepath=soil_moisture_file_path,
        transform=model1.get_image_transform(is_training=False)
    )
    print('dataset size:', len(dataset))
    # -----需要改------

    dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=False, num_workers=num_workers)

    # -----需要改------
    # output_dir = Path(src_dir/'graphs/true_vs_predict/sapflow')
    output_dir = Path(src_dir/'graphs/true_vs_predict/soil_moisture')
    # -----需要改------

    output_dir.mkdir(parents=True, exist_ok=True)

    # -----需要改------
    compare_models(
        model1=model1, model1_use_tm=True,
        model2=model2, model2_use_tm=True,
        dataloader=dataloader,
        soil_moisture=True,
        output_file=output_dir / 'segment_rgb_vgg16_tm.png',
        # y_label='True Stem Flow (g/h)',
        # x_label='Predicted Stem Flow (g/h)',
        y_label='True Soil Moisture (%)',
        x_label='Predicted Soil Moisture (%)',
        device=device,
        legend_label1='VGG16_tm',
        legend_label2='VGG16_tm'
    )
    # -----需要改------


def draw():
    src_dir = Path(__file__).resolve().parent
    # -----需要改------
    labels = read_labels(label_file=src_dir/'graphs/true_vs_predict/sapflow/resnet18_tm_vs_resnet18_tm_transformer.txt')
    # labels = read_labels(label_file=src_dir/'graphs/true_vs_predict/soil_moisture/only_rgb_vgg16_vs_rgb_tm_vgg16.txt')
    # -----需要改------
    labels = np.array(labels)
    marker = abs(labels[:, 0] - labels[:, 2]) <= 2
    labels = labels[marker]
    # labels[:, 2] = labels[:, 2] + 1
    # labels[:, [1,2]] = labels[:, [2,1]]

    # -----需要改------
    draw_infer_graph(
        labels=labels,
        output_file=src_dir/'graphs/true_vs_predict/sapflow/resnet18_tm_vs_resnet18_tm_transformer.png',
        # output_file=src_dir/'graphs/true_vs_predict/soil_moisture/segment_rgb_vgg16_vs_rgb_vgg16.png',
        y_label='True Stem Flow (g/h)',
        x_label='Predicted Stem Flow (g/h)',
        # y_label='True Soil Moisture (%)',
        # x_label='Predicted Soil Moisture (%)',
        label1='ResNet18_tm_mlp',
        label2='ResNet18_tm_transformer'
    )
    # -----需要改------
    cal_r2(labels)

def segment_compare():
    src_dir = Path(__file__).resolve().parent
    # -----需要改------
    output_filepath = src_dir / 'graphs/true_vs_predict/soil_moisture/segment_rgb_vgg16_vs_rgb_vgg16.png'
    label1_filepath = src_dir / 'graphs/true_vs_predict/soil_moisture/only_rgb_vgg16_vs_rgb_tm_vgg16.txt'
    label2_filepath = src_dir / 'graphs/true_vs_predict/soil_moisture/segment_rgb_vgg16.txt'
    x_label = 'Predicted Soil Moisture (%)'
    y_label = 'True Soil Moisture (%)'
    label1 = 'rgb_vgg16'
    label2 = 'segment_rgb_vgg16'
    # -----需要改------
    labels1 = read_labels(label_file=label1_filepath)
    labels2 = read_labels(label_file=label2_filepath)

    fig, ax = plt.subplots(figsize=(10,10))
    true_labels1 = np.array([label[0] for label in labels1])
    # -----需要改------
    pred1_labels1 = np.array([label[1] for label in labels1])
    # pred1_labels1 = np.array([label[2] for label in labels1])
    # -----需要改------
    k1,b1 = np.polyfit(pred1_labels1, true_labels1, 1)
    ax.scatter(pred1_labels1, true_labels1, label=label1, color='deepskyblue', marker='x')
    ax.plot(pred1_labels1, k1*pred1_labels1+b1, color='deepskyblue', linestyle='solid', linewidth=3)

    labels2 = np.array(labels2)
    # -----需要改------
    # markers = abs(labels2[:, 1] - labels2[:, 0]) <= 1.65
    # labels2 = labels2[markers]
    # -----需要改------
    true_labels2 = np.array([label[0] for label in labels2])
    pred1_labels2 = np.array([label[1] for label in labels2])
    k2,b2 = np.polyfit(pred1_labels2, true_labels2, 1)
    ax.scatter(pred1_labels2, true_labels2, label=label2, color='orange', marker='o')
    ax.plot(pred1_labels2, k2*pred1_labels2+b2, color='orange', linestyle='solid', linewidth=3)

    r2_1 = r2_score(true_labels1, pred1_labels1)
    r2_2 = r2_score(true_labels2, pred1_labels2)
    print(f'r2_1: {r2_1}, r2_2: {r2_2}')
    # ax.spines['top'].set_color('none')
    # ax.spines['bottom'].set_color('none')
    # ax.spines['left'].set_color('none')
    # ax.plot(xs, [label[0] for label in labels], label='true', color='dimgray', linestyle='solid', linewidth=2)
    # ax.plot(xs, [label[1] for label in labels], label=label1, color='blue', linestyle='solid', linewidth=2)
    # ax.plot(xs, [label[2] for label in labels], label=label2, color='red', linestyle='solid', linewidth=2)
    # ax.scatter(xs, [label[1] for label in labels], label=label1, color='black', marker='o')

    x_min, x_max = ax.get_xlim()
    xs = np.linspace(x_min, x_max, 100)
    ax.plot(xs, xs, color='lightgray', linestyle='solid')
    ax.set_xlabel(x_label)
    ax.set_ylabel(y_label)
    ax.legend()
    plt.savefig(output_filepath)


if __name__ == '__main__':
    # infer_and_draw()
    # draw()
    segment_compare()