From c874a2e40336dec7ef991ef444a99aa13a073a5d Mon Sep 17 00:00:00 2001
From: Valentyn N Sichkar <34546286+sichkar-valentyn@users.noreply.github.com>
Date: Thu, 16 May 2019 16:04:24 +0300
Subject: [PATCH] Update traffic_sign.md
---
Theory/traffic_sign.md | 90 +++---------------------------------------
1 file changed, 5 insertions(+), 85 deletions(-)
diff --git a/Theory/traffic_sign.md b/Theory/traffic_sign.md
index 6c7aa56..9d78f83 100644
--- a/Theory/traffic_sign.md
+++ b/Theory/traffic_sign.md
@@ -4,8 +4,11 @@ Implementing Traffic Sign Classification in Python for Computer Vision tasks in
## Test online [here](https://valentynsichkar.name/traffic_signs.html)
+### Courses:
+* Explore course **"Convolutional Neural Networks for Image Classification"** here: https://stepik.org/course/53801/promo
+
## Content
-Theory and experimental results (on this page):
+Short description of the content. Full codes you can find inside the course by link above:
* [Loading Data](#loading-data)
* [Unique Training Examples](#unique-examples)
@@ -26,7 +29,7 @@ Theory and experimental results (on this page):
### Loading Data
-Data used for this task is **German Traffic Sign Benchmarks (GTSB)**.
+Data used for this task is **German Traffic Sign Benchmarks (GTSRB)**.
Initially dataset consists of images in ***PPM*** format with different sizes.
For current task datasets were organized in the same way as it was done for [CIFAR-10 Image Classification](https://github.com/sichkar-valentyn/Neural_Networks_for_Computer_Vision/blob/master/Theory/cifar10.md):
@@ -158,8 +161,6 @@ Shapes of data4 - data8 are as following (Gray):
For **Model 1** architecture will be used as it was done for [CIFAR-10 Image Classification](https://github.com/sichkar-valentyn/Neural_Networks_for_Computer_Vision/blob/master/Theory/cifar10.md):
`Input` --> `Conv` --> `ReLU` --> `Pool` --> `Affine` --> `ReLU` --> `Affine` --> `Softmax`
-
-
For **Model 1** following parameters will be used:
| Parameter | Description |
@@ -175,52 +176,6 @@ For **Model 1** architecture will be used as it was done for [CIFAR-10 Image Cla
### Overfitting Small Data for Model 1
For Overfitting Small Data of Model 1 dataset **'data8.pickle'** was chosen.
-
Code for Overfitting Small Data for Model 1 will be used as it was done for [CIFAR-10 Image Classification](https://github.com/sichkar-valentyn/Neural_Networks_for_Computer_Vision/blob/master/Theory/cifar10.md):
-
-```py
-import numpy as np
-
-# Importing module 'ConvNet1.py'
-from Helper_Functions.ConvNet1 import *
-
-# Importing module 'Solver.py'
-from Solver import *
-
-# Loading data
-# Opening file for reading in binary mode
-with open('Data_Preprocessing/data8.pickle', 'rb') as f:
- d = pickle.load(f, encoding='latin1') # dictionary type
-
-# Number of training examples
-number_of_training_data = 10
-
-# Preparing data by slicing in 'data' dictionary appropriate array
-small_data = {
- 'x_train':d['x_train'][:number_of_training_data],
- 'y_train':d['y_train'][:number_of_training_data],
- 'x_validation':d['x_validation'],
- 'y_validation':d['y_validation']
- }
-
-# Creating instance of class for 'ConvNet1' and initializing model
-model = ConvNet1(weight_scale=1e-2, hidden_dimension=100)
-
-# Creating instance of class for 'Solver' and initializing model
-solver = Solver(model,
- small_data,
- update_rule='adam',
- optimization_config={'learning_rate':1e-3},
- learning_rate_decay=1.0,
- batch_size=50,
- number_of_epochs=100,
- print_every=1,
- verbose_mode=True
- )
-
-# Running training process
-solver.train()
-```
-
Overfitting Small Data with 10 training examples and 100 epochs is shown on the figure below.
@@ -229,41 +184,6 @@ Overfitting Small Data with 10 training examples and 100 epochs is shown on the
### Training of Model 1
For training Model 1 dataset **'data8.pickle'** was chosen as it reached the best accuracy over all datasets.
-
Code for training Model 1 will be used as it was done for [CIFAR-10 Image Classification](https://github.com/sichkar-valentyn/Neural_Networks_for_Computer_Vision/blob/master/Theory/cifar10.md):
-
-```py
-import numpy as np
-
-# Importing module 'ConvNet1.py'
-from Helper_Functions.ConvNet1 import *
-
-# Importing module 'Solver.py'
-from Solver import *
-
-# Loading data
-# Opening file for reading in binary mode
-with open('Data_Preprocessing/data8.pickle', 'rb') as f:
- d = pickle.load(f, encoding='latin1') # dictionary type
-
-# Creating instance of class for 'ConvNet1' and initializing model
-model = ConvNet1(input_dimension=(1, 32, 32), weight_scale=1e-3, hidden_dimension=500, regularization=1-e3)
-
-# Creating instance of class for 'Solver' and initializing model
-solver = Solver(model,
- d,
- update_rule='adam',
- optimization_config={'learning_rate':1e-3},
- learning_rate_decay=1.0,
- batch_size=50,
- number_of_epochs=50,
- print_every=1,
- verbose_mode=True
- )
-
-# Running training process
-solver.train()
-```
-
Model 1 with 'data8.pickle' dataset reached **0.989** training accuracy.
Training process of Model 1 with **17 500** iterations is shown on the figure below.