Your task is to implement a single 2D convolutional layer, a core component of Convolutional Neural Networks (CNNs), without relying on deep learning frameworks. This exercise will deepen your understanding of how convolution operations work in computer vision applications.

Task Description

Create a Python function that performs a 2D convolution on an input image using a set of learnable filters. The function should produce feature maps as its output.

Input Parameters

1. input_image: A 3D NumPy array representing a single image (height, width, channels)
2. filters: A 4D NumPy array representing multiple convolutional filters (number of filters, filter height, filter width, input channels)
3. stride: An integer specifying the stride of the convolution (default: 1)
4. padding: A string specifying the padding type ('valid' or 'same', default: 'valid')

Output

• A 3D NumPy array representing the feature maps produced by the convolution operation

Requirements

1. Convolution Operation
   • Implement the forward pass of a 2D convolution
   • Apply multiple filters to the input image
   • Support configurable stride and padding options
2. Activation Function
   • Apply a ReLU (Rectified Linear Unit) activation function to the output
3. Programming Constraints
   • Use Python as the programming language
   • You may use NumPy for array operations
   • Do not use deep learning frameworks like PyTorch or TensorFlow for the convolution operation

Additional Guidelines

• Ensure your implementation handles edge cases properly (e.g., different input sizes, number of channels)
• Optimize your code for efficiency where possible
• Include clear comments explaining your implementation

Bonus Challenges (Optional)

• Implement a backward pass for the convolution operation
• Add support for different activation functions (e.g., sigmoid, tanh)
• Extend the implementation to handle batches of images

Evaluation Criteria

Your implementation will be evaluated based on:

1. Correctness of the convolution operation
2. Proper handling of stride and padding options
3. Correct application of the ReLU activation function
4. Code clarity and efficiency
5. Adherence to Python best practices

Good luck with your implementation!