Building Computer Vision Projects with OpenCV 4 and C++

Training an SVM model

We are now going to use supervised learning and then obtain a set of images for each object and its corresponding label. There is no minimum number of images in the dataset; if we provide more images for the training process, we will get a better classification model (in most cases). However, for simple classifiers, it could be enough to train simple models. To do this, we created three folders (screw, nut, and ring), where all of the images of each type are placed together. For each image in the folder, we have to extract the features, add them to the train feature matrix and, at the same time, create a new vector with the labels for each row corresponding to each training matrix. To evaluate our system, we will split ...