import numpy as np import tensorflow as tf from tensorflow.keras.applications import MobileNetV2 from tensorflow.keras.models import Model from tensorflow.keras.layers import Dense, Flatten, Dropout from pathlib import Path from PIL import Image import csv import os import matplotlib.pyplot as plt from reportlab.pdfgen import canvas from reportlab.lib.pagesizes import A4

#LBL1: Загрузка данных из папки def load_dataset_from_directory(directory: Path): frames, labels, coords = [], [], [] for root, _, files in os.walk(directory): for file in files: if file.endswith('.jpg'): img = Image.open(os.path.join(root, file)).resize((224, 224)) frames.append(np.array(img, dtype=np.uint8)) elif file == 'labels.csv': with open(os.path.join(root, file), newline='') as csvfile: reader = csv.reader(csvfile) next(reader) for row in reader: try: labels.append(int(row[1]) if row[1] else 0) coords.append([int(row[2]) if row[2] else 0, int(row[3]) if row[3] else 0]) except ValueError: print(f"Invalid value in row: {row}") return np.array(frames), np.array(labels), np.array(coords)

#LBL2: Создание модели def initialize_model(): base_model = MobileNetV2(weights='imagenet', include_top=False, input_shape=(224, 224, 3)) base_model.trainable = False x = Dropout(0.5)(Flatten()(base_model.output)) model = Model(inputs=base_model.input, outputs=[Dense(1, activation='sigmoid', name='classification')(x), Dense(2, activation='linear', name='regression')(x)]) return model

Компиляция модели

model = initialize_model() model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=0.001), loss={'classification': 'binary_crossentropy', 'regression': 'mse'}, loss_weights={'classification': 1.0, 'regression': 0.5}, metrics={'classification': 'accuracy'})

#LBL1: Тренировка модели def fit_model(model, train_data, val_data, epochs=10, batch_size=32): train_frames, train_labels, train_coords = train_data val_frames, val_labels, val_coords = val_data train_dataset = tf.data.Dataset.from_tensor_slices( (train_frames, {'classification': train_labels, 'regression': train_coords})).map( lambda x, y: (tf.image.resize(tf.cast(x, tf.float32) / 255.0, (224, 224)), y)).batch(batch_size) val_dataset = tf.data.Dataset.from_tensor_slices( (val_frames, {'classification': val_labels, 'regression': val_coords})).map( lambda x, y: (tf.image.resize(tf.cast(x, tf.float32) / 255.0, (224, 224)), y)).batch(batch_size)

history = model.fit(train_dataset, validation_data=val_dataset, epochs=epochs)

# Создание графиков
plt.figure(figsize=(12, 5))

# График потерь
plt.subplot(1, 2, 1)
plt.plot(history.history['loss'], label='Train Loss')
plt.plot(history.history['val_loss'], label='Val Loss')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.title('Loss Over Epochs')

# График точности
plt.subplot(1, 2, 2)
plt.plot(history.history['classification_accuracy'], label='Train Accuracy')
plt.plot(history.history['val_classification_accuracy'], label='Val Accuracy')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend()
plt.title('Accuracy Over Epochs')

plt.savefig('training_results.png')
plt.show()

return history

#LBL2: Расчет Simple Ball Tracking Accuracy (SiBaTrAcc) def compute_sibatracc(predictions, ground_truth, e1=5, e2=5, step=8, alpha=1.5): def calculate_error(code_pr, x_pr, y_pr, code_gt, x_gt, y_gt): if code_gt != 0 and code_pr == 0: return e1 elif code_gt == 0 and code_pr != 0: return e2 else: distance = np.sqrt((x_gt - x_pr) ** 2 + (y_gt - y_pr) ** 2) return min(5, distance / step) ** alpha

total_error = 0
N = len(predictions)
for pred, gt in