About Files in this repository
1. myData folder
Contains 43 sub folders starting from 0 and ending at 42. Each subfolder has images of relevant classs.
Eg. subfolder named 0 has images of speed limit 20
2. label.csv
This file contains label for each subfolder.
Eg. subfolder 0 has a label "Speed limit (20km/h)" and subfolder 1 has a label "Speed limit (30km/h) and so on
3. model_trained.p
This is a pickle file which stores parameters of trained model that can be reused
4. TraficSign_Main.py
This is the python file that contains code for training the model.
5. TrafficSign_Test.py
This python file contains code to test the model
About Dataset and Model performance
1. Dataset information
Total images - 34791
Color band - 3 (Red, Green, Blue)
Image size - 32 * 32
Total number of classes - 43
Training set size - 22271 images
Test set size - 6960 images (20% of the total dataset)
Validation set size - 5568 images (20% of the remaining dataset(after excluding test set))
2. About Model
Model used - Convolutional Neural Network
Probability Threshold - 75%