Overview

This dataset contains 2093 images of traffic signs in various environments. The goal of this dataset is to provide valuable resources to train custom Computer Vision models capable of identifying the classes listed below.

Model Training

If you'd like to train an Object Detection Model with this dataset, you can follow the step by step guide in our tensorflow-object-detection GitHub repository

Classes and Annotation Format

The dataset includes 2266 annotations across the following object classes:

• bus_stop - 105 instances;
• do_not_enter - 100 instances;
• do_not_stop - 100 instances;
• do_not_turn_l - 113 instances;
• do_not_turn_r - 101 instances;
• do_not_u_turn - 100 instances;
• enter_left_lane - 102 instances;
• green_light - 167 instances;
• left_right_lane - 101 instances;
• no_parking - 104 instances;
• parking - 101 instances;
• ped_crossing - 100 instances;
• ped_zebra_cross - 117 instances;
• railway_crossing - 102 instances;
• red_light - 149 instances;
• stop - 108 instances;
• t_intersection_l - 102 instances;
• traffic_light_sign - 100 instances;
• u_turn - 100 instances;
• warning - 100 instances;
• yellow_light - 94 instances;

The dataset is annotated in the YOLO annotation format, in which each annotation corresponds to a single line in a text file, with each line representing one object instance. The format for each line is as follows:

<object-class> <x_center> <y_center> <width> <height>

• <object-class>: The integer index representing the object's class based on the order they appear in the classes.txt file.

• <x_center>, <y_center>: The normalized coordinates of the center of the bounding box with respect to the width and height of the image. Both <x_center> and <y_center> should be in the range [0, 1]. The top-left corner of the image is (0,0), and the bottom-right corner is (1,1).

• <width>, <height>: The normalized width and height of the bounding box, also relative to the width and height of the image. Both <width> and <height> should be in the range [0, 1].

For any questions or issues related to the dataset, please contact contact@amlstation.com.

🔗 Visit our website: www.amlstation.com

🔗 Follow us on LinkedIn: www.linkedin.com/company/amlstation

License

License: CC BY 4.0

This dataset is based on Traffic Sign Computer Vision Project, which is licensed under a Creative Commons Attribution 4.0 International License.

The following changes were made to the original dataset:

• All images and annotations were placed inside the images folder without being divided into train, test and validation sets;
• Classes names, originally placed inside _darknet.labels files, were placed in classes.txt;

You can find the complete license text at the Creative Commons website: https://creativecommons.org/licenses/by/4.0/