AlexNet C++

A from-scratch C++17 implementation of AlexNet inference using plain tensors and custom layer implementations (Conv2D, ReLU, LRN, MaxPool2D, Linear, Dropout, Softmax).

Project Layout

• include/: layer/model/tensor headers and configuration constants
• src/: implementation files and demo entry point (main.cpp)
• CMakeLists.txt: build configuration

Build

Prerequisites

• CMake 3.16 or newer
• C++17-compatible compiler (AppleClang/Clang/GCC)

Commands

cmake -S . -B build
cmake --build build -j

Run

Run the demo executable:

./build/alexnet_demo

Expected output format:

• Logits shape: [N, 1000]
• First 10 probabilities: ...

The demo currently creates a deterministic synthetic input tensor (1 x 3 x 227 x 227), runs a forward pass, and prints logits/probability summaries.