/SDCND-Term1-Vehicle-Detection-and-Tracking

Primary LanguageJupyter Notebook

Vehicle Detection

Udacity - Self-Driving Car NanoDegree

The goals / steps of this project are the following:

  • Perform a Histogram of Oriented Gradients (HOG) feature extraction on a labeled training set of images and train a classifier Linear SVM classifier
  • Optionally, you can also apply a color transform and append binned color features, as well as histograms of color, to your HOG feature vector.
  • Note: for those first two steps don't forget to normalize your features and randomize a selection for training and testing.
  • Implement a sliding-window technique and use your trained classifier to search for vehicles in images.
  • Run your pipeline on a video stream (start with the test_video.mp4 and later implement on full project_video.mp4) and create a heat map of recurring detections frame by frame to reject outliers and follow detected vehicles.
  • Estimate a bounding box for vehicles detected.

For details on the working of that project checkout the report.md

The following files are important for the project:

  • report.md (report / project writeup)
  • report_files/ (folder with images and videos used for the report)
  • [Vehicle Detection and tracking.ipynb](./Vehicle Detection and tracking.ipynb) (exploration of the algorithms based on single images and 6 frames)
  • test_images/ (folder with images used by the notebook)
  • train_clf.py (the script used to train the classifiers)
  • detect_and_track.py (the script used to annotate the video)
  • helper/*.py (python modules implementing the funcionality of the pipeline)
  • project_video.mp4 (base video for later annotation)
  • result_video.mp4 (final resulting video)

Important: To train the classifier datasets with vehicles and non vehicles are expected to be inside a data folder on the root! Originally the dataset was included, but I could not submitt the project to review with the dataset. I will probably add it after review again.

The video annotation is done with the script detect_and_track.py. It has the values of the input video and the resulting video hardcoded, so if you want to change it, you have to change it in code.