PROJECT TITLE: CLEAN/DIRTY ROAD DETECTION USING TRANSFER LEARNING

Description:

This is a project based on ConvNets used to identify whether a road is clean or dirty. We have used MobileNet as our base architecture and the weights are based on imagenet.

Contents

1. Dataset download.
2. Tools and Libraries.
3. Code.
4. References.

1. Dataset download

We have used web scraping techniques to download 238 images of both clean and dirty roads. This dataset was completely used for training the model.

2. Tools and Libraries

In this project we will be using the Keras Deep Learning Library and we will be running it on top of the Tensorflow backend.
Others include:

• Numpy
• Matplotlib
• OpenCV

3. Code

ClassifierModel.py is the main code which encapsulates the CNN architecture that was used for this project.

• We have used keras.applications.mobilenet_v2 to import the MobileNetV2 architecture for our base model, with training = False.
• On top of it, we have added two dense layers consisting of 512 and 256 units both having ReLU activation followed by Dropout.
• The output layer is a simple 2 unit softmax layer.
  

Training Accuracy and Loss

We have used keras.losses.CategoricalCrossentropy as our loss function, keras.optimizers.Adam as Optimizer and trained the model for 15 epochs.

4. References

• Tensorflow
• Transfer Learning
• Web Scraping
• MobileNet