The dataset shared here is used to train the model for the team playing Goodgame, winning the Digital Race - Contest self-driving car programming for students in 2020. Each team’s task is to train and optimize AI models, and at the same time integrate them on a self-driving car model using NVIDIA’s Jetson TX2 board to follow. in the designated lanes, follow directional signs and avoid obstacles on the road. The Goodgame team, with a lot of talent and effort, won the championship. Goodgame captain Dat Vu helped share the source code and the data you used to train your AI models.

Digital Race 2020

1. Link to download data

The data is divided into two main tasks: Sign detection and lane detection.

• Data for sign detection
• Data for lane segment

2. Information about data

The data for sign detection is structured as follows:

Object Detection
    Data
         |───000000_10.png
         |───000001_10.png
         |───...
    │─── test.csv
    train.csv

This dataset includes 12,764 images in the training set - training set and 2,561 images in the monitoring set - validation set, with 6 classes of signs:

• Turn left
• Turn right
• Go straight
• Stop
• No Turn Left
• No Turn Right

The labels of the dataset are contained in .csv files with the following structure:

Data for sign recognition

Lane Segment Dataset:

Directory structure:

Segmentation
    GGDataSet
         |─── train_frames
             |─── train
                 |─── train_000001.png
         |─── train_masks
             |─── train
                 |─── train_000001.png
         |─── val_frames
             |─── val
                 |─── val_000001.png
         |─── val_masks
             |─── val
                 |─── val_000001.png
         |─── label_colors.txt
    │─── model_pb
    models
    train.py
    │─── convert_pb.py

The data set includes 6240 images in the training set - training set and 1448 images in the monitoring set - validation set, including segment labels for 3 layers: Background, Road lines and Road.

lane segment data

3. Source code

Those interested can see the notebooks for training deep learning networks using these datasets at the following link: https://github.com/makerhanoi/via-dataset. Goodgame also shares their source code to the community at here.