MODEL: Build a Small-CNN for Stop Sign Detection

[obaidmm]

In this task, we will design and implement a Convolutional Neural Network (CNN) for real-time stop sign detection. The CNN is planned to have 7 layers.

The goal is to build a lightweight model that can accurately detect stop signs in images while ensuring it runs efficiently on low-power. The model should be able to process images. To achieve this, we will construct a small CNN with convolutional layers for feature extraction, batch normalization for stable training, pooling layers to reduce spatial dimensions, and fully connected layers for classification (You should do more research on this ). The final layer will use a Sigmoid function to spit out the confidence score of our binary classification : stop sign OR not stop sign. We may use Softmax activation function to output the probability of a stop sign being present if we have more detection classes, i.e. Yield sign, speed limit, etc.

The final deliverable for this task will be a Python script (cnn_detectionmodel.py) that defines the model, loading and preprocesses the dataset, and saves the trained model in an appropriate format (basically .pt).