![]() ![]() Images have been shuffled as well before kicking off the training session. I have divided provided data set into training set (80%) and testing set (20%). HOG features have been scaled to zero mean and unit variance using Standard Scaler. It is a supervised learning model which will be able to classify whether something is a car or not after we train it. For this particular case we will be using Linear Support Vector Machines (Linear SVMs). In order to detect the car based on our feature set, we would need a prediction model. Increasing orientations and pixel per cell parameters did improve prediction time but the accuracy rate of the model went down. HOG feature extraction was based on 9 orientations, 8 pixels per cell and 2 cells per block. Here’s a sample of vehicle and non-vehicle image with HOG features from the same images as above:Įxtracted HOG features from sample training data I have tried other color spaces, but YCrCb gave me the best accuracy when training my prediction model. ![]() I have used YCrCb color space and all its channels as inputs for HOG features extraction. Scikit-image python library provides us with the necessary API for calculating HOG feature. In essence, you should think of features as thumbprints of the objects you are interested in. This feature descriptor is much more resilient to the dynamics of the traffic. In order to have a robust feature set and increase our accuracy rate we will be using Histogram of Oriented Gradients (HOG). We could try using simple template matching or relaying on color features but these methods are not robust enough when it comes to changing perspectives and shapes of the object. In order to detect a car on the image, we need to identify feature(s) which uniquely represent a car. I will definitely go back to this project and try out some of the top performers in this list on the same problem:Įxamples from the training data set Feature extraction With the renaissance of deep learning, convolutional neural networks have automated this task while significantly boosting performance.Īs it turns out, Deep Neural Networks are outperforming the approach which I have used (Linear Support Vector Machines in combination with Histogram of Oriented Gradients). However, all previous methods rely on hand-crafted features that are difficult to design. ![]() With the work of Dalal & Triggs (2005), linear Support Vector Machines (SVMs), that maximizes the margin of all samples from a linear decision boundary, in combination with Histogram of Orientation (HOG) features have become popular tools for classification. The course material is suggesting the usage of somewhat outdated approach for detecting vehicles which I figured out in the middle of the project by reading this great paper on the state-of-the-art computer vision for autonomous vehicles. A snapshot from the final output of the project ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. Archives
January 2023
Categories |