Fan: Boshqaruv tamoyillari Amaliy mashg’ulot Guruh
Yüklə 80,96 Kb.
|
| {_val_f1:.4f}"
f" — val_precision: {_val_precision:.4f}" f" — val_recall: {_val_recall:.4f}"))
In [6]:
batch_size = 64 num_classes = 14 epochs = 30 val_split = 0.1 save_dir = os.path.join(os.getcwd(), 'models') model_name = 'keras_cnn_model.h5' In [7]: model = Sequential() model.add(Conv2D(32, (3, 3), padding='same', input_shape=x_train.shape[1:])) model.add(Activation('relu')) model.add(Conv2D(32, (3, 3))) model.add(Activation('relu')) model.add(MaxPooling2D(pool_size=(2, 2))) model.add(Dropout(0.25))
model.add(Activation('relu')) model.add(Conv2D(64, (3, 3))) model.add(Activation('relu')) model.add(MaxPooling2D(pool_size=(2, 2))) model.add(Dropout(0.25)) model.add(Flatten()) model.add(Dense(512)) model.add(Activation('relu')) model.add(Dropout(0.5)) model.add(Dense(num_classes)) model.add(Activation('softmax')) WARNING:tensorflow:From /opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version. Instructions for updating: Colocations handled automatically by placer. WARNING:tensorflow:From /opt/conda/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version. Instructions for updating: Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`. In [8]: f1_metrics = Metrics() model.compile( loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'] )
x_train, y_train, batch_size=batch_size, epochs=epochs, callbacks=[f1_metrics], validation_split=val_split, shuffle=True ) WARNING:tensorflow:From /opt/conda/lib/python3.6/site-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version. Instructions for updating: Use tf.cast instead. Train on 176669 samples, validate on 19630 samples Epoch 1/30 176669/176669 [==============================] - 25s 139us/step - loss: 0.7065 - acc: 0.7801 - val_loss: 0.4995 - val_acc: 0.8322 val_f1: 0.4737 — val_precision: 0.5971 — val_recall: 0.4506 Epoch 2/30 1088/176669 [..............................] - ETA: 20s - loss: 0.5210 - acc: 0.8346 /opt/conda/lib/python3.6/site-packages/sklearn/metrics/classification.py:1143: UndefinedMetricWarning: F-score is ill-defined and being set to 0.0 in labels with no predicted samples. 'precision', 'predicted', average, warn_for) /opt/conda/lib/python3.6/site-packages/sklearn/metrics/classification.py:1143: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. 'precision', 'predicted', average, warn_for) 176669/176669 [==============================] - 20s 111us/step - loss: 0.5137 - acc: 0.8341 - val_loss: 0.4165 - val_acc: 0.8625 val_f1: 0.6290 — val_precision: 0.7200 — val_recall: 0.5941 Epoch 3/30 176669/176669 [==============================] - 20s 111us/step - loss: 0.4522 - acc: 0.8519 - val_loss: 0.3753 - val_acc: 0.8790 val_f1: 0.7002 — val_precision: 0.7599 — val_recall: 0.6748 Epoch 4/30 176669/176669 [==============================] - 20s 111us/step - loss: 0.4148 - acc: 0.8626 - val_loss: 0.3463 - val_acc: 0.8842 val_f1: 0.7160 — val_precision: 0.7663 — val_recall: 0.6952 Epoch 5/30 176669/176669 [==============================] - 20s 112us/step - loss: 0.3910 - acc: 0.8700 - val_loss: 0.3262 - val_acc: 0.8927 val_f1: 0.7180 — val_precision: 0.7776 — val_recall: 0.6866 Epoch 6/30 176669/176669 [==============================] - 20s 111us/step - loss: 0.3710 - acc: 0.8758 - val_loss: 0.3225 - val_acc: 0.8918 val_f1: 0.7178 — val_precision: 0.7933 — val_recall: 0.6846 Epoch 7/30 176669/176669 [==============================] - 20s 111us/step - loss: 0.3541 - acc: 0.8809 - val_loss: 0.2957 - val_acc: 0.9013 val_f1: 0.7680 — val_precision: 0.7901 — val_recall: 0.7544 Epoch 8/30 176669/176669 [==============================] - 20s 113us/step - loss: 0.3408 - acc: 0.8855 - val_loss: 0.2802 - val_acc: 0.9071 val_f1: 0.7833 — val_precision: 0.8065 — val_recall: 0.7684 Epoch 9/30 176669/176669 [==============================] - 20s 111us/step - loss: 0.3303 - acc: 0.8886 - val_loss: 0.2747 - val_acc: 0.9090 val_f1: 0.7855 — val_precision: 0.8201 — val_recall: 0.7706 Epoch 10/30 176669/176669 [==============================] - 20s 111us/step - loss: 0.3210 - acc: 0.8912 - val_loss: 0.2724 - val_acc: 0.9096 val_f1: 0.7905 — val_precision: 0.8167 — val_recall: 0.7755 Epoch 11/30 176669/176669 [==============================] - 20s 112us/step - loss: 0.3152 - acc: 0.8930 - val_loss: 0.2702 - val_acc: 0.9121 val_f1: 0.7730 — val_precision: 0.8145 — val_recall: 0.7514 Epoch 12/30 176669/176669 [==============================] - 20s 111us/step - loss: 0.3073 - acc: 0.8956 - val_loss: 0.2615 - val_acc: 0.9138 val_f1: 0.7628 — val_precision: 0.7943 — val_recall: 0.7451 Epoch 13/30 176669/176669 [==============================] - 20s 111us/step - loss: 0.3013 - acc: 0.8981 - val_loss: 0.2598 - val_acc: 0.9150 val_f1: 0.8018 — val_precision: 0.8323 — val_recall: 0.7893 Epoch 14/30 176669/176669 [==============================] - 20s 111us/step - loss: 0.2966 - acc: 0.8988 - val_loss: 0.2506 - val_acc: 0.9163 val_f1: 0.8082 — val_precision: 0.8147 — val_recall: 0.8059 Epoch 15/30 176669/176669 [==============================] - 20s 111us/step - loss: 0.2893 - acc: 0.9017 - val_loss: 0.2577 - val_acc: 0.9157 val_f1: 0.8010 — val_precision: 0.8308 — val_recall: 0.7830 Epoch 16/30 176669/176669 [==============================] - 20s 111us/step - loss: 0.2838 - acc: 0.9026 - val_loss: 0.2418 - val_acc: 0.9213 val_f1: 0.8145 — val_precision: 0.8275 — val_recall: 0.8065 Epoch 17/30 176669/176669 [==============================] - 20s 111us/step - loss: 0.2814 - acc: 0.9035 - val_loss: 0.2419 - val_acc: 0.9213 val_f1: 0.8157 — val_precision: 0.8357 — val_recall: 0.8044 Epoch 18/30 12224/176669 [=>............................] - ETA: 17s - loss: 0.2751 - acc: 0.9074 In [9]:
if not os.path.isdir(save_dir): os.makedirs(save_dir) model_path = os.path.join(save_dir, model_name) model.save(model_path) print('Saved trained model at %s ' % model_path) Saved trained model at /kaggle/working/models/keras_cnn_model.h5 In [10]: history_df = pd.DataFrame(hist.history) history_df['val_f1'] = f1_metrics.val_f1s history_df['val_precision'] = f1_metrics.val_precisions history_df['val_recall'] = f1_metrics.val_recalls
history_df[['acc', 'val_acc']].plot() history_df[['val_f1', 'val_precision', 'val_recall']].plot() Out[10]:
In [11]: y_test = model.predict(x_test) submission_df = pd.read_csv('../input/iwildcam-2019-fgvc6/sample_submission.csv') submission_df['Predicted'] = y_test.argmax(axis=1) print(submission_df.shape) submission_df.head() (153730, 2) Out[11]:
In [12]: submission_df.to_csv('submission.csv',index=False) history_df.to_csv('history.csv', index=False) with open('history.json', 'w') as f: json.dump(hist.history, f) Yüklə 80,96 Kb. Dostları ilə paylaş:
|