0. Intro
There are a lot of people who want to get into Google's Deep Learning library, Tensorflow, but the hardest part of the first person is probably having to install the Linux operating system to use the GPU. I was doing all the development in the existing Windows environment, but it was hard to come up with a new HW for the new Linux installation. In addition, the fact that I was not familiar with the Linux environment was also a difficult part. So I installed Linux in a virtual environment such as VMware or Virtual Box and installed it up to Tensorflow, and confirmed the operation. The biggest reason for not being able to make progress was that the GPU did not work in a virtual environment and it took too much time to learn.
So I tried various experiments using MXnet among the libraries that can use Deep Learning on Windows instead of Tensorflow, but since there are not many users, I did not have much the latest technology implementation sources than Tensorflow.
In the meantime, on November 29, 2016, Tensorflow v0.12.0 RC0 was released, and a library that fully supports GPU on Windows was announced! Cheer up! Finally I can do Tensorflow !!
So what do I do with Tensorflow? As a result, I thought that I would try to apply image classification to my images. I would like to post all the results obtained by shoveling for 2 weeks. In particular, the model used for learning was the slim included in Tensorflow. When I first used slim, I could not find the material posted in Korea.
Based on this posting, you can also use it immediately in the field where you want to apply image classification.
Posting Order
0. Intro
1. Installing the Tensorflow GPU version in Windows
2. Download, learn and evaluate slim models
3. Learning from my images (using caltech images)
4. Using learned models
5. Creating a Python Tkinter GUI application
So let's get started.
댓글 없음:
댓글 쓰기