File size: 12,884 Bytes
08d66df
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
# ============================================================
# SMARTVISION AI - MODEL 1 (v2): VGG16 (TRANSFER LEARNING + FT)
# with proper preprocess_input + label smoothing + deeper FT
# ============================================================

import os
import time
import json
import numpy as np
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers

from sklearn.metrics import (
    precision_recall_fscore_support,
    confusion_matrix,
    classification_report,
)

from tensorflow.keras.applications.vgg16 import VGG16, preprocess_input

print("TensorFlow version:", tf.__version__)

# ------------------------------------------------------------
# 1. CONFIGURATION
# ------------------------------------------------------------

BASE_DIR      = "smartvision_dataset"  # your dataset root
CLASS_DIR     = os.path.join(BASE_DIR, "classification")
TRAIN_DIR     = os.path.join(CLASS_DIR, "train")
VAL_DIR       = os.path.join(CLASS_DIR, "val")
TEST_DIR      = os.path.join(CLASS_DIR, "test")

IMG_SIZE      = (224, 224)
BATCH_SIZE    = 32
NUM_CLASSES   = 25

MODELS_DIR    = "saved_models"
METRICS_DIR   = "smartvision_metrics"

os.makedirs(MODELS_DIR, exist_ok=True)
os.makedirs(METRICS_DIR, exist_ok=True)

print("Train dir:", TRAIN_DIR)
print("Val dir  :", VAL_DIR)
print("Test dir :", TEST_DIR)

# ------------------------------------------------------------
# 2. LOAD DATASETS (FROM CROPPED SINGLE-OBJECT IMAGES)
# ------------------------------------------------------------

train_ds = tf.keras.utils.image_dataset_from_directory(
    TRAIN_DIR,
    image_size=IMG_SIZE,
    batch_size=BATCH_SIZE,
    shuffle=True,
)

val_ds = tf.keras.utils.image_dataset_from_directory(
    VAL_DIR,
    image_size=IMG_SIZE,
    batch_size=BATCH_SIZE,
    shuffle=False,
)

test_ds = tf.keras.utils.image_dataset_from_directory(
    TEST_DIR,
    image_size=IMG_SIZE,
    batch_size=BATCH_SIZE,
    shuffle=False,
)

class_names = train_ds.class_names
print("Detected classes:", class_names)
print("Number of classes:", len(class_names))

AUTOTUNE = tf.data.AUTOTUNE
train_ds = train_ds.prefetch(AUTOTUNE)
val_ds   = val_ds.prefetch(AUTOTUNE)
test_ds  = test_ds.prefetch(AUTOTUNE)

# ------------------------------------------------------------
# 3. DATA AUGMENTATION (APPLIED ONLY DURING TRAINING)
# ------------------------------------------------------------

data_augmentation = keras.Sequential(
    [
        layers.RandomFlip("horizontal"),                 # random horizontal flips
        layers.RandomRotation(0.04),                     # ≈ ±15 degrees
        layers.RandomZoom(0.1),                          # random zoom
        layers.RandomContrast(0.2),                      # ±20% contrast
        layers.Lambda(lambda x: tf.image.random_brightness(x, max_delta=0.2)),
        layers.Lambda(lambda x: tf.image.random_saturation(x, 0.8, 1.2)),
    ],
    name="data_augmentation",
)

# NOTE:
# We DO NOT use Rescaling(1./255) here.
# Instead, we use VGG16's preprocess_input which subtracts ImageNet means
# and expects BGR ordering. This matches the pretrained weights.

# ------------------------------------------------------------
# 4. BUILD VGG16 MODEL (FROZEN BASE + CUSTOM HEAD)
# ------------------------------------------------------------

def build_vgg16_model_v2():
    inputs = keras.Input(shape=(*IMG_SIZE, 3), name="input_layer")

    # 1. Augmentation (only active during training)
    x = data_augmentation(inputs)

    # 2. VGG16-specific preprocessing
    x = layers.Lambda(
        lambda z: preprocess_input(tf.cast(z, tf.float32)),
        name="vgg16_preprocess"
    )(x)

    # 3. Pre-trained VGG16 backbone (no top classification head)
    base_model = VGG16(
        include_top=False,
        weights="imagenet",
        input_tensor=x,
    )

    # Freeze backbone initially (Stage 1)
    base_model.trainable = False

    # 4. Custom classification head for 25 classes
    x = layers.GlobalAveragePooling2D(name="global_average_pooling2d")(base_model.output)
    x = layers.Dense(256, activation="relu", name="dense_256")(x)
    x = layers.Dropout(0.5, name="dropout_0_5")(x)
    outputs = layers.Dense(NUM_CLASSES, activation="softmax", name="predictions")(x)

    model = keras.Model(inputs=inputs, outputs=outputs, name="VGG16_smartvision_v2")
    return model

vgg16_model = build_vgg16_model_v2()
vgg16_model.summary()

# ------------------------------------------------------------
# 5. CUSTOM LOSS WITH LABEL SMOOTHING
# ------------------------------------------------------------

def make_sparse_ce_with_label_smoothing(num_classes, label_smoothing=0.05):
    """
    Implements sparse categorical crossentropy with manual label smoothing.
    Works even if your Keras version doesn't support `label_smoothing` in
    SparseCategoricalCrossentropy.__init__.
    """
    ls = float(label_smoothing)
    nc = int(num_classes)

    def loss_fn(y_true, y_pred):
        # y_true: integer labels, shape (batch,)
        y_true = tf.cast(y_true, tf.int32)
        y_true_oh = tf.one_hot(y_true, depth=nc)

        if ls > 0.0:
            smooth = ls
            y_true_oh = (1.0 - smooth) * y_true_oh + smooth / tf.cast(nc, tf.float32)

        # y_pred is softmax probabilities
        return tf.keras.losses.categorical_crossentropy(
            y_true_oh, y_pred, from_logits=False
        )

    return loss_fn

# ------------------------------------------------------------
# 6. TRAINING UTILITY (COMMON FOR STAGE 1 & 2)
# ------------------------------------------------------------

def compile_and_train(
    model,
    model_name,
    train_ds,
    val_ds,
    epochs,
    lr,
    model_tag,
    patience_es=5,
    patience_rlr=2,
):
    """
    Compile and train model, saving the best weights by val_accuracy.
    model_name: base name ("vgg16_v2")
    model_tag : "stage1" or "stage2" etc.
    """
    print(f"\n===== TRAINING {model_name} ({model_tag}) =====")

    optimizer = keras.optimizers.Adam(learning_rate=lr)

    # Use our custom loss with label smoothing
    loss_fn = make_sparse_ce_with_label_smoothing(
        num_classes=NUM_CLASSES,
        label_smoothing=0.05,
    )

    model.compile(
        optimizer=optimizer,
        loss=loss_fn,
        metrics=["accuracy"],
    )

    best_weights_path = os.path.join(MODELS_DIR, f"{model_name}_{model_tag}_best.h5")

    callbacks = [
        keras.callbacks.ModelCheckpoint(
            filepath=best_weights_path,
            monitor="val_accuracy",
            save_best_only=True,
            mode="max",
            verbose=1,
        ),
        keras.callbacks.EarlyStopping(
            monitor="val_accuracy",
            patience=patience_es,
            restore_best_weights=True,
            verbose=1,
        ),
        keras.callbacks.ReduceLROnPlateau(
            monitor="val_loss",
            factor=0.5,
            patience=patience_rlr,
            min_lr=1e-6,
            verbose=1,
        ),
    ]

    history = model.fit(
        train_ds,
        validation_data=val_ds,
        epochs=epochs,
        callbacks=callbacks,
    )

    return history, best_weights_path

# ------------------------------------------------------------
# 7. STAGE 1: TRAIN HEAD WITH FROZEN VGG16 BASE
# ------------------------------------------------------------

print("\n===== STAGE 1: Training head with frozen VGG16 base =====")

# Safety: ensure all VGG16 conv blocks are frozen
for layer in vgg16_model.layers:
    if layer.name.startswith("block"):
        layer.trainable = False

epochs_stage1 = 20
lr_stage1     = 1e-4

history_stage1, vgg16_stage1_best = compile_and_train(
    vgg16_model,
    model_name="vgg16_v2",
    train_ds=train_ds,
    val_ds=val_ds,
    epochs=epochs_stage1,
    lr=lr_stage1,
    model_tag="stage1",
    patience_es=5,
    patience_rlr=2,
)

print("Stage 1 best weights saved at:", vgg16_stage1_best)

# ------------------------------------------------------------
# 8. STAGE 2: FINE-TUNE BLOCK4 + BLOCK5 OF VGG16
# ------------------------------------------------------------

print("\n===== STAGE 2: Fine-tuning VGG16 block4 + block5 =====")

# Load best Stage 1 weights before fine-tuning
vgg16_model.load_weights(vgg16_stage1_best)

# Unfreeze only block4_* and block5_* layers for controlled fine-tuning
for layer in vgg16_model.layers:
    if layer.name.startswith("block5") :
        layer.trainable = True      # fine-tune top two blocks
    elif layer.name.startswith("block"):
        layer.trainable = False     # keep lower blocks frozen (block1–3)

# Head layers (GAP + Dense + Dropout + output) remain trainable

epochs_stage2 = 15
lr_stage2     = 1e-5   # slightly higher than 1e-5 but still safe for FT

history_stage2, vgg16_stage2_best = compile_and_train(
    vgg16_model,
    model_name="vgg16_v2",
    train_ds=train_ds,
    val_ds=val_ds,
    epochs=epochs_stage2,
    lr=lr_stage2,
    model_tag="stage2",
    patience_es=6,
    patience_rlr=3,
)

print("Stage 2 best weights saved at:", vgg16_stage2_best)

# ------------------------------------------------------------
# 9. EVALUATION + SAVE METRICS & CONFUSION MATRIX
# ------------------------------------------------------------

def evaluate_and_save(model, model_name, best_weights_path, test_ds, class_names):
    print(f"\n===== EVALUATING {model_name.upper()} ON TEST SET =====")

    # Load best weights
    model.load_weights(best_weights_path)
    print(f"Loaded best weights from {best_weights_path}")

    y_true = []
    y_pred = []
    all_probs = []

    total_time = 0.0
    total_images = 0

    # Predict over test dataset
    for images, labels in test_ds:
        images_np = images.numpy()
        bs = images_np.shape[0]

        start = time.perf_counter()
        probs = model.predict(images_np, verbose=0)
        end = time.perf_counter()

        total_time += (end - start)
        total_images += bs

        preds = np.argmax(probs, axis=1)

        y_true.extend(labels.numpy())
        y_pred.extend(preds)
        all_probs.append(probs)

    y_true = np.array(y_true)
    y_pred = np.array(y_pred)
    all_probs = np.concatenate(all_probs, axis=0)

    # Basic metrics
    accuracy = float((y_true == y_pred).mean())
    precision, recall, f1, _ = precision_recall_fscore_support(
        y_true, y_pred, average="weighted", zero_division=0
    )

    # Top-5 accuracy
    top5_correct = 0
    for i, label in enumerate(y_true):
        if label in np.argsort(all_probs[i])[-5:]:
            top5_correct += 1
    top5_acc = top5_correct / len(y_true)

    # Inference time
    time_per_image = total_time / total_images
    images_per_second = 1.0 / time_per_image

    # Model size (weights only)
    temp_w = os.path.join(MODELS_DIR, f"{model_name}_temp_for_size.weights.h5")
    model.save_weights(temp_w)
    size_mb = os.path.getsize(temp_w) / (1024 * 1024)
    os.remove(temp_w)

    # Confusion matrix
    cm = confusion_matrix(y_true, y_pred)

    print("\nClassification Report:")
    print(classification_report(y_true, y_pred, target_names=class_names, zero_division=0))

    print(f"Test Accuracy        : {accuracy:.4f}")
    print(f"Weighted Precision   : {precision:.4f}")
    print(f"Weighted Recall      : {recall:.4f}")
    print(f"Weighted F1-score    : {f1:.4f}")
    print(f"Top-5 Accuracy       : {top5_acc:.4f}")
    print(f"Avg time per image   : {time_per_image*1000:.2f} ms")
    print(f"Images per second    : {images_per_second:.2f}")
    print(f"Model size (weights) : {size_mb:.2f} MB")
    print(f"Num parameters       : {model.count_params()}")

    # Save metrics + confusion matrix in dedicated folder
    save_dir = os.path.join(METRICS_DIR, model_name)
    os.makedirs(save_dir, exist_ok=True)

    metrics = {
        "model_name": model_name,
        "accuracy": accuracy,
        "precision_weighted": float(precision),
        "recall_weighted": float(recall),
        "f1_weighted": float(f1),
        "top5_accuracy": float(top5_acc),
        "avg_inference_time_sec": float(time_per_image),
        "images_per_second": float(images_per_second),
        "model_size_mb": float(size_mb),
        "num_parameters": int(model.count_params()),
    }

    metrics_path = os.path.join(save_dir, "metrics.json")
    cm_path = os.path.join(save_dir, "confusion_matrix.npy")

    with open(metrics_path, "w") as f:
        json.dump(metrics, f, indent=2)

    np.save(cm_path, cm)

    print(f"\nSaved metrics to        : {metrics_path}")
    print(f"Saved confusion matrix to: {cm_path}")

    return metrics, cm


# Evaluate FINAL (fine-tuned) model on test set
vgg16_metrics, vgg16_cm = evaluate_and_save(
    vgg16_model,
    model_name="vgg16_v2_stage2",
    best_weights_path=vgg16_stage2_best,
    test_ds=test_ds,
    class_names=class_names,
)

print("\n✅ VGG16 v2 (2-stage, improved) pipeline complete.")