Skip to content

Traffic Sign Alert is a Deep Learning project where Deep Neural Networks have been used to classify, recognise and label Traffic Signs. The final model has been tested to perform at 96% accuracy for 12K testing images.

License

Notifications You must be signed in to change notification settings

SamarpanDas/Traffic-Sign-Alert

Repository files navigation

Traffic-Sign-Alert-

Under Developmental stage

Link to download the actual dataset : https://www.kaggle.com/meowmeowmeowmeowmeow/gtsrb-german-traffic-sign

The Training set was ideal to use, it had subfolders 0 - 42 (which means a total of 43 classes) containing images for its respective class.

The Test set had 12,000 images(approx) randomly arranged with a "test.csv" file attached which had 8 columns among which there was a column conatining the imagePath and its corresponding classId.

I found the original test data set uncomfortable to use, thus I wrote the "Support for Traffic (Initial Test set file Handling).ipynb" file which created a Testset folder and subsequently created 43 subfolders(named from 0 - 42) inside it. Then the following lines of code picked up images from the original Test Set and put those images in the various subfolders in the newly created Test Set as per their classId mentioned in the test.csv file.

Therefore the original training set and the newly created test set were used to train and evaluate the Model which is present in the TrafficMain.ipynb file. The train and test set used now resembled the same format.

Train/Test set ---> 43 subfolders (named from 0 - 42 ; referring to the 43 classes) ---> containing their respective images.

About

Traffic Sign Alert is a Deep Learning project where Deep Neural Networks have been used to classify, recognise and label Traffic Signs. The final model has been tested to perform at 96% accuracy for 12K testing images.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published