Skip to content
Snippets Groups Projects
Commit 26ceb97d authored by Jean-Luc Parouty's avatar Jean-Luc Parouty
Browse files

update GTSRB

parent 9f16240a
No related branches found
No related tags found
No related merge requests found
Hide whitespace changes
Inline Side-by-side
.gitignore
+ 1
− 1
View file @ 26ceb97d
...@@ -2,6 +2,6 @@ ...@@ -2,6 +2,6 @@
*/.ipynb_checkpoints/* */.ipynb_checkpoints/*
__pycache__ __pycache__
*/__pycache__/* */__pycache__/*
/run/** run/
*/data/* */data/*
!/GTSRB/data/dataset.tar.gz !/GTSRB/data/dataset.tar.gz
GTSRB/03-Tracking-and-visualizing.ipynb
+ 86
− 61
View file @ 26ceb97d
%% Cell type:markdown id: tags: %% Cell type:markdown id: tags:
German Traffic Sign Recognition Benchmark (GTSRB) German Traffic Sign Recognition Benchmark (GTSRB)
================================================= =================================================
--- ---
Introduction au Deep Learning (IDLE) - S. Aria, E. Maldonado, JL. Parouty - CNRS/SARI/DEVLOG - 2020 Introduction au Deep Learning (IDLE) - S. Aria, E. Maldonado, JL. Parouty - CNRS/SARI/DEVLOG - 2020
## Episode 3 : Tracking and visualizing ## Episode 3 : Tracking and visualizing
Our main steps: Our main steps:
- Monitoring and understanding our model training - Monitoring and understanding our model training
- Analyze the results - Analyze the results
- Improving our model - Improving our model
- Add recovery points - Add recovery points
## 1/ Import and init ## 1/ Import and init
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
import tensorflow as tf import tensorflow as tf
from tensorflow import keras from tensorflow import keras
from tensorflow.keras.callbacks import TensorBoard from tensorflow.keras.callbacks import TensorBoard
import numpy as np import numpy as np
import matplotlib import matplotlib
import matplotlib.pyplot as plt import matplotlib.pyplot as plt
import time import time
from importlib import reload from importlib import reload
import idle.pwk as ooo import idle.pwk as ooo
ooo.init() ooo.init()
``` ```
%% Output %% Output
IDLE 2020 - Practical Work Module IDLE 2020 - Practical Work Module
Version : 0.1.1 Version : 0.1.1
Run time : Monday 6 January 2020, 20:52:54 Run time : Monday 6 January 2020, 23:42:11
Matplotlib style : idle/talk.mplstyle Matplotlib style : idle/talk.mplstyle
TensorFlow version : 2.0.0 TensorFlow version : 2.0.0
Keras version : 2.2.4-tf Keras version : 2.2.4-tf
%% Cell type:markdown id: tags: %% Cell type:markdown id: tags:
## 2/ Reload dataset ## 2/ Reload dataset
Dataset is one of the saved dataset: RGB25, RGB35, L25, L35, etc. Dataset is one of the saved dataset: RGB25, RGB35, L25, L35, etc.
First of all, we're going to use the dataset : **L25** First of all, we're going to use the dataset : **L25**
(with a GPU, it only takes 35'' compared to more than 5' with a CPU !) (with a GPU, it only takes 35'' compared to more than 5' with a CPU !)
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
%%time %%time
dataset ='L25' dataset ='L25'
img_lx = 25 img_lx = 25
img_ly = 25 img_ly = 25
img_lz = 1 img_lz = 1
# ---- Read dataset # ---- Read dataset
x_train = np.load('./data/{}/x_train.npy'.format(dataset)) x_train = np.load('./data/{}/x_train.npy'.format(dataset))
y_train = np.load('./data/{}/y_train.npy'.format(dataset)) y_train = np.load('./data/{}/y_train.npy'.format(dataset))
x_test = np.load('./data/{}/x_test.npy'.format(dataset)) x_test = np.load('./data/{}/x_test.npy'.format(dataset))
y_test = np.load('./data/{}/y_test.npy'.format(dataset)) y_test = np.load('./data/{}/y_test.npy'.format(dataset))
# ---- Reshape data # ---- Reshape data
x_train = x_train.reshape( x_train.shape[0], img_lx, img_ly, img_lz) x_train = x_train.reshape( x_train.shape[0], img_lx, img_ly, img_lz)
x_test = x_test.reshape( x_test.shape[0], img_lx, img_ly, img_lz) x_test = x_test.reshape( x_test.shape[0], img_lx, img_ly, img_lz)
input_shape = (img_lx, img_ly, img_lz) input_shape = (img_lx, img_ly, img_lz)
print("Dataset loaded, size={:.1f} Mo\n".format(ooo.get_directory_size('./data/'+dataset))) print("Dataset loaded, size={:.1f} Mo\n".format(ooo.get_directory_size('./data/'+dataset)))
``` ```
%% Output %% Output
Dataset loaded, size=247.6 Mo Dataset loaded, size=247.6 Mo
CPU times: user 0 ns, sys: 297 ms, total: 297 ms CPU times: user 0 ns, sys: 344 ms, total: 344 ms
Wall time: 330 ms Wall time: 498 ms
%% Cell type:markdown id: tags: %% Cell type:markdown id: tags:
## 3/ Have a look to the dataset ## 3/ Have a look to the dataset
Note: Data must be reshape for matplotlib Note: Data must be reshape for matplotlib
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
print("x_train : ", x_train.shape) print("x_train : ", x_train.shape)
print("y_train : ", y_train.shape) print("y_train : ", y_train.shape)
print("x_test : ", x_test.shape) print("x_test : ", x_test.shape)
print("y_test : ", y_test.shape) print("y_test : ", y_test.shape)
if img_lz>1: if img_lz>1:
ooo.plot_images(x_train.reshape(-1,img_lx,img_ly,img_lz), y_train, range(6), columns=3, x_size=4, y_size=3) ooo.plot_images(x_train.reshape(-1,img_lx,img_ly,img_lz), y_train, range(6), columns=3, x_size=4, y_size=3)
ooo.plot_images(x_train.reshape(-1,img_lx,img_ly,img_lz), y_train, range(36), columns=12, x_size=1, y_size=1) ooo.plot_images(x_train.reshape(-1,img_lx,img_ly,img_lz), y_train, range(36), columns=12, x_size=1, y_size=1)
else: else:
ooo.plot_images(x_train.reshape(-1,img_lx,img_ly), y_train, range(6), columns=6, x_size=2, y_size=2) ooo.plot_images(x_train.reshape(-1,img_lx,img_ly), y_train, range(6), columns=6, x_size=2, y_size=2)
ooo.plot_images(x_train.reshape(-1,img_lx,img_ly), y_train, range(36), columns=12, x_size=1, y_size=1) ooo.plot_images(x_train.reshape(-1,img_lx,img_ly), y_train, range(36), columns=12, x_size=1, y_size=1)
``` ```
%% Output %% Output
x_train : (39209, 25, 25, 1) x_train : (39209, 25, 25, 1)
y_train : (39209,) y_train : (39209,)
x_test : (12630, 25, 25, 1) x_test : (12630, 25, 25, 1)
y_test : (12630,) y_test : (12630,)
%% Cell type:markdown id: tags: %% Cell type:markdown id: tags:
## 4/ Create model ## 4/ Create model
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
batch_size = 64 batch_size = 64
num_classes = 43 num_classes = 43
epochs = 20 epochs = 10
``` ```
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
tf.keras.backend.clear_session()
model = keras.models.Sequential() model = keras.models.Sequential()
model.add( keras.layers.Conv2D(96, (3,3), activation='relu', input_shape=(img_lx, img_ly, img_lz))) model.add( keras.layers.Conv2D(96, (3,3), activation='relu', input_shape=(img_lx, img_ly, img_lz)))
model.add( keras.layers.MaxPooling2D((2, 2))) model.add( keras.layers.MaxPooling2D((2, 2)))
model.add( keras.layers.Conv2D(192, (3, 3), activation='relu')) model.add( keras.layers.Conv2D(192, (3, 3), activation='relu'))
model.add( keras.layers.MaxPooling2D((2, 2))) model.add( keras.layers.MaxPooling2D((2, 2)))
model.add( keras.layers.Flatten()) model.add( keras.layers.Flatten())
model.add( keras.layers.Dense(3072, activation='relu')) model.add( keras.layers.Dense(3072, activation='relu'))
model.add( keras.layers.Dense(500, activation='relu')) model.add( keras.layers.Dense(500, activation='relu'))
model.add( keras.layers.Dense(500, activation='relu')) model.add( keras.layers.Dense(500, activation='relu'))
model.add( keras.layers.Dense(43, activation='softmax')) model.add( keras.layers.Dense(43, activation='softmax'))
model.summary() model.summary()
model.compile(optimizer='adam', model.compile(optimizer='adam',
loss='sparse_categorical_crossentropy', loss='sparse_categorical_crossentropy',
metrics=['accuracy']) metrics=['accuracy'])
``` ```
%% Output %% Output
Model: "sequential" Model: "sequential"
_________________________________________________________________ _________________________________________________________________
Layer (type) Output Shape Param # Layer (type) Output Shape Param #
================================================================= =================================================================
conv2d (Conv2D) (None, 23, 23, 96) 960 conv2d (Conv2D) (None, 23, 23, 96) 960
_________________________________________________________________ _________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 11, 11, 96) 0 max_pooling2d (MaxPooling2D) (None, 11, 11, 96) 0
_________________________________________________________________ _________________________________________________________________
conv2d_1 (Conv2D) (None, 9, 9, 192) 166080 conv2d_1 (Conv2D) (None, 9, 9, 192) 166080
_________________________________________________________________ _________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 4, 4, 192) 0 max_pooling2d_1 (MaxPooling2 (None, 4, 4, 192) 0
_________________________________________________________________ _________________________________________________________________
flatten (Flatten) (None, 3072) 0 flatten (Flatten) (None, 3072) 0
_________________________________________________________________ _________________________________________________________________
dense (Dense) (None, 3072) 9440256 dense (Dense) (None, 3072) 9440256
_________________________________________________________________ _________________________________________________________________
dense_1 (Dense) (None, 500) 1536500 dense_1 (Dense) (None, 500) 1536500
_________________________________________________________________ _________________________________________________________________
dense_2 (Dense) (None, 500) 250500 dense_2 (Dense) (None, 500) 250500
_________________________________________________________________ _________________________________________________________________
dense_3 (Dense) (None, 43) 21543 dense_3 (Dense) (None, 43) 21543
================================================================= =================================================================
Total params: 11,415,839 Total params: 11,415,839
Trainable params: 11,415,839 Trainable params: 11,415,839
Non-trainable params: 0 Non-trainable params: 0
_________________________________________________________________ _________________________________________________________________
%% Cell type:markdown id: tags: %% Cell type:markdown id: tags:
## 5/ Add callbacks ## 5/ Add callbacks
Nous allons ajouter 2 callbacks : Nous allons ajouter 2 callbacks :
- **TensorBoard** - **TensorBoard**
Training logs, which can be visualised with Tensorboard. Training logs, which can be visualised with Tensorboard.
`#tensorboard --logdir ./run/logs` `#tensorboard --logdir ./run/logs`
- model backup IMPORTANT : Relancer tensorboard à chaque run
- **model backup**
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
# reload(ooo) # ---- Callback tensorboard
# ---- Callback for tensorboard log_dir = "./run/logs/" + ooo.tag_now()
log_dir="./run/logs/" + ooo.tag_now()
tensorboard_callback = tf.keras.callbacks.TensorBoard(log_dir=log_dir, histogram_freq=1) tensorboard_callback = tf.keras.callbacks.TensorBoard(log_dir=log_dir, histogram_freq=1)
# ---- Callback ModelCheckpoint
save_dir = "./run/models"
checkpoint_callback = tf.keras.callbacks.ModelCheckpoint(filepath=save_dir, verbose=0, monitor='accuracy', save_best_only=True)
``` ```
%% Cell type:markdown id: tags: %% Cell type:markdown id: tags:
## 5/ Run model ## 5/ Run model
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
%%time %%time
history = model.fit( x_train[:3000], y_train[:3000], history = model.fit( x_train[:3000], y_train[:3000],
batch_size=batch_size, batch_size=batch_size,
epochs=epochs, epochs=epochs,
verbose=1, verbose=1,
validation_data=(x_test, y_test), validation_data=(x_test[:500], y_test[:500]),
callbacks=[tensorboard_callback] ) callbacks=[tensorboard_callback, checkpoint_callback] )
``` ```
%% Output %% Output
Train on 3000 samples, validate on 12630 samples Train on 3000 samples, validate on 500 samples
Epoch 1/20 Epoch 1/10
3000/3000 [==============================] - 12s 4ms/sample - loss: 3.4839 - accuracy: 0.0740 - val_loss: 3.1409 - val_accuracy: 0.1995 2944/3000 [============================>.] - ETA: 0s - loss: 3.4893 - accuracy: 0.0676INFO:tensorflow:Assets written to: ./run/models/assets
Epoch 2/20 3000/3000 [==============================] - 8s 3ms/sample - loss: 3.4847 - accuracy: 0.0690 - val_loss: 3.2739 - val_accuracy: 0.1640
3000/3000 [==============================] - 12s 4ms/sample - loss: 2.4676 - accuracy: 0.3067 - val_loss: 2.1162 - val_accuracy: 0.3968 Epoch 2/10
Epoch 3/20 2944/3000 [============================>.] - ETA: 0s - loss: 2.4907 - accuracy: 0.3227INFO:tensorflow:Assets written to: ./run/models/assets
3000/3000 [==============================] - 13s 4ms/sample - loss: 1.4458 - accuracy: 0.5543 - val_loss: 1.4468 - val_accuracy: 0.5862 3000/3000 [==============================] - 7s 2ms/sample - loss: 2.4802 - accuracy: 0.3250 - val_loss: 2.0143 - val_accuracy: 0.3900
Epoch 4/20 Epoch 3/10
3000/3000 [==============================] - 14s 5ms/sample - loss: 0.9291 - accuracy: 0.7067 - val_loss: 1.1903 - val_accuracy: 0.6622 2944/3000 [============================>.] - ETA: 0s - loss: 1.3810 - accuracy: 0.5591INFO:tensorflow:Assets written to: ./run/models/assets
Epoch 5/20 3000/3000 [==============================] - 7s 2ms/sample - loss: 1.3794 - accuracy: 0.5593 - val_loss: 1.3322 - val_accuracy: 0.6200
3000/3000 [==============================] - 14s 5ms/sample - loss: 0.6027 - accuracy: 0.8030 - val_loss: 0.8067 - val_accuracy: 0.7900 Epoch 4/10
Epoch 6/20 2944/3000 [============================>.] - ETA: 0s - loss: 0.8501 - accuracy: 0.7286INFO:tensorflow:Assets written to: ./run/models/assets
3000/3000 [==============================] - 14s 5ms/sample - loss: 0.3835 - accuracy: 0.8670 - val_loss: 0.8453 - val_accuracy: 0.7925 3000/3000 [==============================] - 7s 2ms/sample - loss: 0.8426 - accuracy: 0.7323 - val_loss: 1.1705 - val_accuracy: 0.6580
Epoch 7/20 Epoch 5/10
3000/3000 [==============================] - 14s 5ms/sample - loss: 0.2666 - accuracy: 0.9157 - val_loss: 0.7578 - val_accuracy: 0.8256 2944/3000 [============================>.] - ETA: 0s - loss: 0.5132 - accuracy: 0.8400INFO:tensorflow:Assets written to: ./run/models/assets
Epoch 8/20 3000/3000 [==============================] - 8s 3ms/sample - loss: 0.5087 - accuracy: 0.8413 - val_loss: 0.9281 - val_accuracy: 0.7360
3000/3000 [==============================] - 14s 5ms/sample - loss: 0.2045 - accuracy: 0.9307 - val_loss: 0.8074 - val_accuracy: 0.8300 Epoch 6/10
Epoch 9/20 2944/3000 [============================>.] - ETA: 0s - loss: 0.3740 - accuracy: 0.8787INFO:tensorflow:Assets written to: ./run/models/assets
3000/3000 [==============================] - 14s 5ms/sample - loss: 0.1370 - accuracy: 0.9567 - val_loss: 0.7071 - val_accuracy: 0.8588 3000/3000 [==============================] - 8s 3ms/sample - loss: 0.3763 - accuracy: 0.8783 - val_loss: 0.9252 - val_accuracy: 0.7520
Epoch 10/20 Epoch 7/10
3000/3000 [==============================] - 12s 4ms/sample - loss: 0.0964 - accuracy: 0.9707 - val_loss: 0.7275 - val_accuracy: 0.8622 2944/3000 [============================>.] - ETA: 0s - loss: 0.2671 - accuracy: 0.9141INFO:tensorflow:Assets written to: ./run/models/assets
Epoch 11/20 3000/3000 [==============================] - 8s 3ms/sample - loss: 0.2677 - accuracy: 0.9140 - val_loss: 0.8153 - val_accuracy: 0.8000
3000/3000 [==============================] - 13s 4ms/sample - loss: 0.0654 - accuracy: 0.9803 - val_loss: 0.7073 - val_accuracy: 0.8661 Epoch 8/10
Epoch 12/20 2944/3000 [============================>.] - ETA: 0s - loss: 0.1599 - accuracy: 0.9535INFO:tensorflow:Assets written to: ./run/models/assets
3000/3000 [==============================] - 13s 4ms/sample - loss: 0.0806 - accuracy: 0.9757 - val_loss: 0.8265 - val_accuracy: 0.8637 3000/3000 [==============================] - 8s 3ms/sample - loss: 0.1588 - accuracy: 0.9543 - val_loss: 0.7000 - val_accuracy: 0.8500
Epoch 13/20 Epoch 9/10
3000/3000 [==============================] - 13s 4ms/sample - loss: 0.0854 - accuracy: 0.9743 - val_loss: 0.7547 - val_accuracy: 0.8635 2944/3000 [============================>.] - ETA: 0s - loss: 0.1282 - accuracy: 0.9626INFO:tensorflow:Assets written to: ./run/models/assets
Epoch 14/20 3000/3000 [==============================] - 8s 3ms/sample - loss: 0.1271 - accuracy: 0.9633 - val_loss: 0.7833 - val_accuracy: 0.8080
3000/3000 [==============================] - 14s 5ms/sample - loss: 0.0463 - accuracy: 0.9860 - val_loss: 0.7248 - val_accuracy: 0.8846 Epoch 10/10
Epoch 15/20 2944/3000 [============================>.] - ETA: 0s - loss: 0.0979 - accuracy: 0.9715INFO:tensorflow:Assets written to: ./run/models/assets
3000/3000 [==============================] - 14s 5ms/sample - loss: 0.0425 - accuracy: 0.9917 - val_loss: 0.7266 - val_accuracy: 0.8785 3000/3000 [==============================] - 8s 3ms/sample - loss: 0.0971 - accuracy: 0.9720 - val_loss: 0.7634 - val_accuracy: 0.8420
Epoch 16/20 CPU times: user 4min 26s, sys: 1min 31s, total: 5min 58s
3000/3000 [==============================] - 12s 4ms/sample - loss: 0.0322 - accuracy: 0.9923 - val_loss: 0.7501 - val_accuracy: 0.8777 Wall time: 1min 15s
Epoch 17/20
3000/3000 [==============================] - 14s 5ms/sample - loss: 0.0375 - accuracy: 0.9870 - val_loss: 0.7782 - val_accuracy: 0.8811
Epoch 18/20
3000/3000 [==============================] - 14s 5ms/sample - loss: 0.0557 - accuracy: 0.9857 - val_loss: 0.6846 - val_accuracy: 0.8760
Epoch 19/20
3000/3000 [==============================] - 14s 5ms/sample - loss: 0.0100 - accuracy: 0.9987 - val_loss: 0.7676 - val_accuracy: 0.8873
Epoch 20/20
3000/3000 [==============================] - 14s 5ms/sample - loss: 0.0025 - accuracy: 1.0000 - val_loss: 0.7481 - val_accuracy: 0.8993
CPU times: user 17min 9s, sys: 5min 55s, total: 23min 4s
Wall time: 4min 29s
%% Cell type:markdown id: tags: %% Cell type:markdown id: tags:
## 6/ Evaluation ## 6/ Evaluation
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
score = model.evaluate(x_test, y_test, verbose=0) score = model.evaluate(x_test, y_test, verbose=0)
print('Test loss : {:5.4f}'.format(score[0])) print('Test loss : {:5.4f}'.format(score[0]))
print('Test accuracy : {:5.4f}'.format(score[1])) print('Test accuracy : {:5.4f}'.format(score[1]))
``` ```
%% Cell type:markdown id: tags: %% Cell type:markdown id: tags:
## 7/ History ## 7/ History
The return of model.fit() returns us the learning history The return of model.fit() returns us the learning history
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
ooo.plot_history(history) ooo.plot_history(history)
``` ```
%% Cell type:markdown id: tags: %% Cell type:markdown id: tags:
--- ---
### Results : ### Results :
L25 : size=250 Mo, 93.15% L25 : size=250 Mo, 93.15%
... ...
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
!ls -l {save_dir}
!tree {save_dir}
``` ```
%% Output
total 176
drwxr-xr-x 1 pjluc pjluc 512 Jan 7 00:16 assets
-rw-r--r-- 1 pjluc pjluc 168427 Jan 7 00:17 saved_model.pb
drwxr-xr-x 1 pjluc pjluc 512 Jan 7 00:17 variables
./run/models
├── assets
├── saved_model.pb
└── variables
├── variables.data-00000-of-00001
└── variables.index
2 directories, 3 files
%% Cell type:markdown id: tags:
A suivre : https://www.tensorflow.org/tutorials/keras/save_and_load
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment