Traffic Light Detector

Welcome to the Traffic Light Detector Repository by Oscar ROSAS (PF Lab @ The University of Tokyo) In this repository you will find a python project that quickly enables you to start working on the Traffic Light Detection problem. The project documentation is available in this link.

For any discussions, comments, or questions, you can contact me directly.

Getting Started

The main components for you to get the project up and running are the ones below:

  • Anaconda 3

  • Python 3

In addition, the project assumes that you have the following configuration on your machine:

  • A Linux-based OS distribution

  • An Nvidia architecture GPU device

How to Install

Using the terminal, execute the following instructions

  1. Install Conda Virtual Environment Manager (the example installs Miniconda, but you can use Anaconda as well)

    $ wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh && bash Miniconda3-latest-Linux-x86_64.sh
$ conda activate base && conda update conda

  2. Clone this repository

    $ git clone https://github.com/oskr27/traffic_light traffic_light

  3. Create the virtual environment from the YML file in the repository

    $ cd traffic_light && conda env create -f traffic_light/environment.yml

  4. Activate the virtual environment, and you should be all set

    $ conda activate traffic

Testing a single image

After you cloned the repository and ensuring all the dependencies are met, you may want to infer the class of a single image. For this, execute the instructions below in a terminal.

$ cd traffic_lights
$ python test_inference.py '../data/training-dataset/inference/dayClip1_154.jpg'

Output:

'../data/training-dataset/inference/dayClip1_154.jpg','../data/training-dataset/models/state_dict/resnet-34-no_tuning_dict.pth','go',0.950,0.166[ms]
'../data/training-dataset/inference/dayClip1_154.jpg','../data/training-dataset/models/state_dict/resnet-50-no_tuning_dict.pth','go',0.963,0.00719[ms]
'../data/training-dataset/inference/dayClip1_154.jpg','../data/training-dataset/models/state_dict/densenet-121-no_tuning_dict.pth','go',0.915,0.0164[ms]
'../data/training-dataset/inference/dayClip1_154.jpg','../data/training-dataset/models/state_dict/resnet-34-tuned_dict.pth','go',0.996,0.00518[ms]
'../data/training-dataset/inference/dayClip1_154.jpg','../data/training-dataset/models/state_dict/resnet-50-tuned_dict.pth','go',0.959,0.00658[ms]
'../data/training-dataset/inference/dayClip1_154.jpg','../data/training-dataset/models/state_dict/densenet-121-tuned_dict.pth','go',0.937,0.0126[ms]

Format: Image_path, Model_path, inference_result, class_score, inference_time