#4 From CNN Training to EfficientNet, Let’s Talk About OpenCV

In my previous post, I shared some work on training a Convolutional Neural Network (CNN) to classify images. Next, I’ll be moving on to try EfficientNet-B2, which usually gives better results.
But before diving into that, let’s talk about OpenCV — because it’s one of the most important tools I used in this project.

OpenCV (Open Source Computer Vision Library) is a powerful library for image processing and computer vision. It allows you to handle tasks such as reading images, converting between color spaces, resizing, or even advanced computer vision tasks like object detection and tracking. In practice, OpenCV is what makes it possible to prepare raw image data before feeding it into a deep learning model.

In my case, I used OpenCV mainly to:

Read images from my dataset.

Convert them to RGB, since models usually expect this format instead of BGR.

Resize them to a consistent size (260×260), so they match the input shape required by the neural network.

This preprocessing step might seem simple, but it’s crucial: deep learning models can only perform well if the input data is clean and consistent.

The code snippet below shows how I combined OpenCV with TensorFlow/Keras to create a pipeline: organizing images into folders, preprocessing them, and training a CNN model for classification.