Week 3 — Emotion Detection
Hello everyone! We continue to provide information about the progress of our Machine Learning project. In this post, we will give detailed information about the dataset we will use in our project. What is the size of the dataset? Is it sufficient for training? Are the images useful? What is the size of the images? How many classes are in the dataset? How many images are in each class? You can be assured that you will find the answer to your questions in the rest of the post. Let’s start!
Our data set is divided into train and validation. We have approximately 30k images in the Train section. In the validation section, there are 6k images. As we explained before, our images consist of 7 classes. To put it briefly, these classes are angry, disgust, fear, happy, neutral, sad, surprised.
In the train section of our data, we have 4k, 436,4k, 7k, 5k, 5k, 3k data respectively. In the validation section, we have 960,111,1k, 2k, 1k, 1k, 797 images respectively. Also, all images in our data set are designated as grayscale.
There are some problems that may be a problem for us. To illustrate these, some of the images are meaningless in our dataset. This image above is an example of this problem for the fear class. Another problem is that the size of our images is 48 * 48. This problem can create problems for us in the algorithm we will establish in the CNN section. Below are sample images for some classes. Because the size of the images is small, the resolution is reduced when we enlarge them.
We also plan to get good results because our data set is large. We are very enthusiastic about the progress of the project :)
We will continue to explain the progress next post. Let’s meet again next week!
