3D CNN Forward Pass Implementation
Core Vision Operations DS practice problem on Onlearn.
Difficulty: medium.
Topics: Understanding 3D Convolutional Neural Network Forward Pass Operations, 3D Cross-correlation, Receptive Field Calculation, Output Dimension Inference, Dot Product Accumulation, Spatial Dimensionality Reduction, Deep Learning, Linear Algebra, Computer Vision, Signal Processing, Computational Complexity, Tensor Operations, Feature Extraction, Weight Sharing, Kernel Sliding Windows, Multi-dimensional Arrays.
Implement a forward pass function for a 3D Convolutional layer. Given a 3D input tensor of shape (D, H, W) and a 3D kernel of shape (kD, kH, kW), compute the output volume assuming a stride of 1 and no padding. The input and kernel are 3D matrices represented as nested lists.