Beyond MNIST Example: Practical Convolutional NNs
What you'll learn
- Enrolled students can easily code all the subprocesses of convolutional neural network.
- Enrolled students can distinguish the problem of not-working Convolutional Neural Networks.
- Enrolled students can apply several solutions for the loading images on Python.
- Enrolled students can understand and apply transfer learning.
- Enrolled students can diagnose imbalanced dataset problem.
- Enrolled students can solve imbalanced dataset problem with different methods.
- Enrolled students can apply several solutions for the image augmentation.
- Enrolled students can create their CustomDataGenerator Functions for Keras.
Requirements
- Python
- Entry Level of Knowledge of Convolutional Neural Networks
- Entry Level of Knowledge of Keras
Description
From the owner of Makine Öğrenmesi (Machine Learning) channel on YouTube that has 4500+ subscribers that is nearly active for 2 years.
makineogrenmesi
Hi. I am Burak, Industrial Engineering student from Bilkent University. I'm working on Machine Learning/Deep Learning for two years and I have an experience on education/practise of Convolutional Neural Network.
Nearly 700,000 minutes of Watch Time. Nearly 170,000 view. Have been 700+ conversion on their problems with subscribers.
I know problems you can face with. My plan is to make you familiar with problems under my control to learn better.
You probably started to learn convolutional neural network with MNIST Tutorial, which is a good example.
However, there are some untold mysteries about it. When you tried to apply CNN to your dataset, you probably had problems that you do not know. What are these problems that can be seen when we take the lid off(easy to hard):
Image Size
Data Size
Arranging the Data
Loading Images
Preparing Data (with several techniques)
Imbalanced Dataset
Several solutions to Imbalanced Dataset
Small Data
Solutions to Small Data
Image Augmentation (Easily handling it with Keras)
Extreme Image Augmentation using dedicated libraries that can be implemented easily on Keras' fit functions.
This is not a course that I only talk about concepts briefly and code it once. All methods are the methods that you can use for general CNN problems.
Also, we will be in direct contact in Udemy Platform. Every advises and problems will be considered by me. So you aren't only registering for course, you can contact with me for any case of deep learning. I will try to help you about the concepts/codes that we are interested in the course.
Who this course is for:
- Deep Learning Enthusiasts that has trouble going one more step after MNIST example and programmers who need practice on using Python libraries that are directly/indirectly related to Deep Learning Libraries such as Tensorflow, PyTorch, Keras
- Python beginners who works on Python for Deep Learning and Machine Learning
- If you need harsh image augmentation, you should be here.
Instructor
I am Burak from Turkey. I am a third year undergraduate student at Bilkent University Industrial Engineering. I am working on Machine Learning and Deep Learning for the last two years. I noticed the lack of Turkish resources in ML/DL at that time and started a YouTube channel (now Turkish language deficiency situation got smaller, lots of blogs opened). I'm trying to learn and teach ML/DL algorithms and applications on youtube / makineogrenmesi . I currently have 4000 followers and people:
Practical Deep Learning:
Deep Learning with Keras:
Applied Machine Learning:
Machine Learning:
hear about the topics such as Python, SciKit, NumPy, Keras. The reason I opened up a separate series called "Practical", aim is to teach people the tricky parts. There are lots of Youtube channels that doesn't talk about the hard part of the coding and finishing the tutorials in five minutes, I'm focusing on searching for solutions to the problems that I have not encountered before and presenting to the beginners. In addition, I have the autonomous car project on GTA Vice City that you can find on Youtube.
After this introduction, I got the right to participate in ISMAIL2017, an artificial intelligence event at Boğaziçi University where graduate and doctoral students were accepted, and I prepared the data in the autonomous car project according to the principles of "Striving for Simplicity: The All Convolutional Net" As a demo, I entered the technical details and got the opportunity to present this work to the participating 50-60 persons.
Since last year we have been working on ML / DL applications in micromachining with my Professor at Bilkent University.