Skip to content

Latest commit

 

History

History
34 lines (23 loc) · 1.39 KB

README.md

File metadata and controls

34 lines (23 loc) · 1.39 KB

Single Shot Detector on Vehicle Dataset

Problem

  • Given Deep learning problem for self-driving cars by training a Single Shot Detector (SSD) to recognize both the front and rear views of vehicles using a vehicle dataset curated and labeled by Davis King of dlib.

  • Each image in this dataset was captured from a camera mounted to a car’s dashboard. For each image, all visible front and rear views of vehicles are labeled as the image below.

  • Given the dlib vehicel dataset (the dataset can be downloaded from here link). The task is to use SSD to detect and localize front and rear of cars

    example

Built With

  • Python
  • Tensorflow Object Detection API (TFOD API)
  • Google Colab

Approach

  • The vehicles dataset was labeled by XML file so first we need to extract the class labels by using BeautifulSoup library
  • Then convert all images and its labels to tensorflow record file
  • Use pre-trained model in this project I used SSD inception v1 for tf 1.x in order to use this step we need to clone the TFOD API the link can be found here

Training process over 200,000 steps, which took approximately 13 hours on Google Colab. The loss ended at approximately 1.7

Visualize the loss and accuracy in TensorBoard

Loss

Accuracy

Result

1 2 3