From 3f9304f3056a102b7efa931e729258811066bc07 Mon Sep 17 00:00:00 2001
From: "Jean-Luc Parouty Jean-Luc.Parouty@simap.grenoble-inp.fr"
<paroutyj@bigfoot3.u-ga.fr>
Date: Thu, 9 Jan 2020 00:40:26 +0100
Subject: [PATCH] Test models on GPU
---
GTSRB/02-First-convolutions.ipynb | 144 ++++++++++++++++++------------
1 file changed, 86 insertions(+), 58 deletions(-)
diff --git a/GTSRB/02-First-convolutions.ipynb b/GTSRB/02-First-convolutions.ipynb
index 3730973..20ee6ca 100644
--- a/GTSRB/02-First-convolutions.ipynb
+++ b/GTSRB/02-First-convolutions.ipynb
@@ -34,7 +34,7 @@
"text": [
"IDLE 2020 - Practical Work Module\n",
" Version : 0.1.1\n",
- " Run time : Wednesday 8 January 2020, 17:47:24\n",
+ " Run time : Wednesday 8 January 2020, 23:42:33\n",
" Matplotlib style : idle/talk.mplstyle\n",
" TensorFlow version : 2.0.0\n",
" Keras version : 2.2.4-tf\n"
@@ -69,7 +69,7 @@
},
{
"cell_type": "code",
- "execution_count": 77,
+ "execution_count": 2,
"metadata": {},
"outputs": [
{
@@ -78,8 +78,8 @@
"text": [
"Dataset loaded, size=911.6 Mo\n",
"\n",
- "CPU times: user 0 ns, sys: 360 ms, total: 360 ms\n",
- "Wall time: 362 ms\n"
+ "CPU times: user 0 ns, sys: 460 ms, total: 460 ms\n",
+ "Wall time: 6.21 s\n"
]
}
],
@@ -117,7 +117,7 @@
},
{
"cell_type": "code",
- "execution_count": 78,
+ "execution_count": 3,
"metadata": {},
"outputs": [
{
@@ -174,18 +174,18 @@
},
{
"cell_type": "code",
- "execution_count": 79,
+ "execution_count": 106,
"metadata": {},
"outputs": [],
"source": [
"batch_size = 64\n",
"num_classes = 43\n",
- "epochs = 5"
+ "epochs = 16"
]
},
{
"cell_type": "code",
- "execution_count": 89,
+ "execution_count": 107,
"metadata": {},
"outputs": [],
"source": [
@@ -229,23 +229,24 @@
"\n",
"def get_model_v3():\n",
" model = keras.models.Sequential()\n",
- " model.add( keras.layers.Conv2D(96, (3,3), activation='relu', padding='same', input_shape=(img_lx, img_ly, img_lz)))\n",
+ " model.add( keras.layers.Conv2D(32, (3,3), activation='relu', input_shape=(img_lx, img_ly, img_lz)))\n",
" model.add( keras.layers.MaxPooling2D((2, 2)))\n",
- " model.add( keras.layers.Dropout(0.2))\n",
+ " model.add( keras.layers.Dropout(0.5))\n",
"\n",
- " model.add( keras.layers.Conv2D(192, (3, 3), activation='relu', padding='same'))\n",
+ " model.add( keras.layers.Conv2D(64, (3, 3), activation='relu'))\n",
" model.add( keras.layers.MaxPooling2D((2, 2)))\n",
- " model.add( keras.layers.Dropout(0.2))\n",
+ " model.add( keras.layers.Dropout(0.5))\n",
"\n",
- " model.add( keras.layers.Conv2D(256, (3, 3), activation='relu', padding='same'))\n",
+ " model.add( keras.layers.Conv2D(128, (3, 3), activation='relu'))\n",
" model.add( keras.layers.MaxPooling2D((2, 2)))\n",
- " model.add( keras.layers.Dropout(0.2))\n",
+ " model.add( keras.layers.Dropout(0.5))\n",
"\n",
- " model.add( keras.layers.Flatten()) \n",
- " model.add( keras.layers.Dense(512, activation='relu'))\n",
- " model.add( keras.layers.Dropout(0.2))\n",
+ " model.add( keras.layers.Conv2D(256, (3, 3), activation='relu'))\n",
+ " model.add( keras.layers.MaxPooling2D((2, 2)))\n",
+ " model.add( keras.layers.Dropout(0.5))\n",
"\n",
- " model.add( keras.layers.Dense(512, activation='relu'))\n",
+ " model.add( keras.layers.Flatten()) \n",
+ " model.add( keras.layers.Dense(1152, activation='relu'))\n",
" model.add( keras.layers.Dropout(0.5))\n",
"\n",
" model.add( keras.layers.Dense(43, activation='softmax'))\n",
@@ -254,55 +255,59 @@
},
{
"cell_type": "code",
- "execution_count": 90,
+ "execution_count": 111,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
- "Model: \"sequential_19\"\n",
+ "Model: \"sequential\"\n",
"_________________________________________________________________\n",
"Layer (type) Output Shape Param # \n",
"=================================================================\n",
- "conv2d_63 (Conv2D) (None, 48, 48, 96) 960 \n",
+ "conv2d (Conv2D) (None, 46, 46, 32) 320 \n",
+ "_________________________________________________________________\n",
+ "max_pooling2d (MaxPooling2D) (None, 23, 23, 32) 0 \n",
"_________________________________________________________________\n",
- "max_pooling2d_48 (MaxPooling (None, 24, 24, 96) 0 \n",
+ "dropout (Dropout) (None, 23, 23, 32) 0 \n",
"_________________________________________________________________\n",
- "dropout_31 (Dropout) (None, 24, 24, 96) 0 \n",
+ "conv2d_1 (Conv2D) (None, 21, 21, 64) 18496 \n",
"_________________________________________________________________\n",
- "conv2d_64 (Conv2D) (None, 24, 24, 192) 166080 \n",
+ "max_pooling2d_1 (MaxPooling2 (None, 10, 10, 64) 0 \n",
"_________________________________________________________________\n",
- "max_pooling2d_49 (MaxPooling (None, 12, 12, 192) 0 \n",
+ "dropout_1 (Dropout) (None, 10, 10, 64) 0 \n",
"_________________________________________________________________\n",
- "dropout_32 (Dropout) (None, 12, 12, 192) 0 \n",
+ "conv2d_2 (Conv2D) (None, 8, 8, 128) 73856 \n",
"_________________________________________________________________\n",
- "conv2d_65 (Conv2D) (None, 12, 12, 256) 442624 \n",
+ "max_pooling2d_2 (MaxPooling2 (None, 4, 4, 128) 0 \n",
"_________________________________________________________________\n",
- "max_pooling2d_50 (MaxPooling (None, 6, 6, 256) 0 \n",
+ "dropout_2 (Dropout) (None, 4, 4, 128) 0 \n",
"_________________________________________________________________\n",
- "dropout_33 (Dropout) (None, 6, 6, 256) 0 \n",
+ "conv2d_3 (Conv2D) (None, 2, 2, 256) 295168 \n",
"_________________________________________________________________\n",
- "flatten_19 (Flatten) (None, 9216) 0 \n",
+ "max_pooling2d_3 (MaxPooling2 (None, 1, 1, 256) 0 \n",
"_________________________________________________________________\n",
- "dense_65 (Dense) (None, 512) 4719104 \n",
+ "dropout_3 (Dropout) (None, 1, 1, 256) 0 \n",
"_________________________________________________________________\n",
- "dropout_34 (Dropout) (None, 512) 0 \n",
+ "flatten (Flatten) (None, 256) 0 \n",
"_________________________________________________________________\n",
- "dense_66 (Dense) (None, 512) 262656 \n",
+ "dense (Dense) (None, 1152) 296064 \n",
"_________________________________________________________________\n",
- "dropout_35 (Dropout) (None, 512) 0 \n",
+ "dropout_4 (Dropout) (None, 1152) 0 \n",
"_________________________________________________________________\n",
- "dense_67 (Dense) (None, 43) 22059 \n",
+ "dense_1 (Dense) (None, 43) 49579 \n",
"=================================================================\n",
- "Total params: 5,613,483\n",
- "Trainable params: 5,613,483\n",
+ "Total params: 733,483\n",
+ "Trainable params: 733,483\n",
"Non-trainable params: 0\n",
"_________________________________________________________________\n"
]
}
],
"source": [
+ "tf.keras.backend.clear_session()\n",
+ "\n",
"model = get_model_v3()\n",
"\n",
"model.summary()\n",
@@ -321,7 +326,7 @@
},
{
"cell_type": "code",
- "execution_count": 91,
+ "execution_count": 112,
"metadata": {},
"outputs": [
{
@@ -329,18 +334,40 @@
"output_type": "stream",
"text": [
"Train on 39209 samples, validate on 12630 samples\n",
- "Epoch 1/5\n",
- "39209/39209 [==============================] - 6s 160us/sample - loss: 1.2276 - accuracy: 0.6518 - val_loss: 0.1684 - val_accuracy: 0.9474\n",
- "Epoch 2/5\n",
- "39209/39209 [==============================] - 6s 140us/sample - loss: 0.1075 - accuracy: 0.9666 - val_loss: 0.1226 - val_accuracy: 0.9625\n",
- "Epoch 3/5\n",
- "39209/39209 [==============================] - 5s 140us/sample - loss: 0.0603 - accuracy: 0.9819 - val_loss: 0.1489 - val_accuracy: 0.9572\n",
- "Epoch 4/5\n",
- "39209/39209 [==============================] - 5s 139us/sample - loss: 0.0465 - accuracy: 0.9855 - val_loss: 0.1102 - val_accuracy: 0.9687\n",
- "Epoch 5/5\n",
- "39209/39209 [==============================] - 5s 137us/sample - loss: 0.0394 - accuracy: 0.9874 - val_loss: 0.0998 - val_accuracy: 0.9727\n",
- "CPU times: user 24.7 s, sys: 3.58 s, total: 28.3 s\n",
- "Wall time: 28.1 s\n"
+ "Epoch 1/16\n",
+ "39209/39209 [==============================] - 3s 81us/sample - loss: 2.4208 - accuracy: 0.3102 - val_loss: 0.7642 - val_accuracy: 0.7483\n",
+ "Epoch 2/16\n",
+ "39209/39209 [==============================] - 3s 66us/sample - loss: 0.7524 - accuracy: 0.7564 - val_loss: 0.3062 - val_accuracy: 0.9110\n",
+ "Epoch 3/16\n",
+ "39209/39209 [==============================] - 3s 66us/sample - loss: 0.4660 - accuracy: 0.8507 - val_loss: 0.2012 - val_accuracy: 0.9437\n",
+ "Epoch 4/16\n",
+ "39209/39209 [==============================] - 3s 66us/sample - loss: 0.3529 - accuracy: 0.8866 - val_loss: 0.1578 - val_accuracy: 0.9536\n",
+ "Epoch 5/16\n",
+ "39209/39209 [==============================] - 3s 65us/sample - loss: 0.2962 - accuracy: 0.9050 - val_loss: 0.1439 - val_accuracy: 0.9614\n",
+ "Epoch 6/16\n",
+ "39209/39209 [==============================] - 3s 67us/sample - loss: 0.2676 - accuracy: 0.9170 - val_loss: 0.1240 - val_accuracy: 0.9660\n",
+ "Epoch 7/16\n",
+ "39209/39209 [==============================] - 3s 66us/sample - loss: 0.2362 - accuracy: 0.9248 - val_loss: 0.1272 - val_accuracy: 0.9667\n",
+ "Epoch 8/16\n",
+ "39209/39209 [==============================] - 3s 67us/sample - loss: 0.2242 - accuracy: 0.9303 - val_loss: 0.1183 - val_accuracy: 0.9671\n",
+ "Epoch 9/16\n",
+ "39209/39209 [==============================] - 3s 65us/sample - loss: 0.2118 - accuracy: 0.9365 - val_loss: 0.1093 - val_accuracy: 0.9709\n",
+ "Epoch 10/16\n",
+ "39209/39209 [==============================] - 3s 66us/sample - loss: 0.1984 - accuracy: 0.9395 - val_loss: 0.1124 - val_accuracy: 0.9725\n",
+ "Epoch 11/16\n",
+ "39209/39209 [==============================] - 3s 64us/sample - loss: 0.1950 - accuracy: 0.9404 - val_loss: 0.1098 - val_accuracy: 0.9712\n",
+ "Epoch 12/16\n",
+ "39209/39209 [==============================] - 3s 65us/sample - loss: 0.1882 - accuracy: 0.9432 - val_loss: 0.1068 - val_accuracy: 0.9727\n",
+ "Epoch 13/16\n",
+ "39209/39209 [==============================] - 3s 65us/sample - loss: 0.1856 - accuracy: 0.9442 - val_loss: 0.1172 - val_accuracy: 0.9702\n",
+ "Epoch 14/16\n",
+ "39209/39209 [==============================] - 3s 66us/sample - loss: 0.1760 - accuracy: 0.9468 - val_loss: 0.1249 - val_accuracy: 0.9694\n",
+ "Epoch 15/16\n",
+ "39209/39209 [==============================] - 3s 66us/sample - loss: 0.1732 - accuracy: 0.9469 - val_loss: 0.1091 - val_accuracy: 0.9719\n",
+ "Epoch 16/16\n",
+ "39209/39209 [==============================] - 3s 65us/sample - loss: 0.1628 - accuracy: 0.9500 - val_loss: 0.1026 - val_accuracy: 0.9743\n",
+ "CPU times: user 52.5 s, sys: 4.32 s, total: 56.8 s\n",
+ "Wall time: 41.9 s\n"
]
}
],
@@ -363,15 +390,15 @@
},
{
"cell_type": "code",
- "execution_count": 92,
+ "execution_count": 110,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
- "Test loss : 0.0998\n",
- "Test accuracy : 0.9727\n"
+ "Test loss : 0.1051\n",
+ "Test accuracy : 0.9770\n"
]
}
],
@@ -388,15 +415,16 @@
"source": [
"---\n",
"### Results : \n",
+ "```\n",
"set-24x24-L : size=230 Mo, 90.67% \n",
"set-24x24-L-LHE : size=230 Mo, 93.90% \n",
"set-24x24-RGB : size=784 Mo, 92.82% \n",
"set-24x24-RGB-HE : size=784 Mo, 92.64% \n",
- "set-48x48-L-LHE : size=230 Mo, 97.21% (v2)\n",
- "set-48x48-RGB-HE : size=xxx Mo, 96.94% (v2)\n",
- "set-48x48-L-LHE : size=784 Mo, 97.24% (v3)\n",
- "\n",
- "..."
+ "set-48x48-L-LHE : size=230 Mo, 97.70% (v2) 1'52\" \n",
+ "set-48x48-RGB-HE : size=xxx Mo, 96.94% (v2) \n",
+ "set-48x48-L-LHE : size=784 Mo, 97.67% (v3) 42\" \n",
+ "...\n",
+ "```"
]
},
{
--
GitLab