Car Detection with Artificial Neural Network

Rear-view vehicle detection using artificial neural network and hog features.

Method

The project was made by the following steps:

1. Feature extraction
2. Network Training
3. Sliding Windows
4. Image Pyramid
5. Detection filtering
6. Results

1. Feature extraction

The select features are HOG.

2. Network Training

The project dataset consist in 9391 Vehicles images and 8493 non vechicle images, all of them in a size of 64x64 pixels.

The dataset was a combination of the both GTI and KITTI Open datasets.

3. Sliding Windows

4. Image Pyramid

5. Detection filtering

This proccess was made by filter in the following steps

ANN values

Heatmap by Overlay

6. Results