Rico's Nerd Cluster

「离开世界之前 一切都是过程」

C++ - Macros

Why Macros, Compiler Args

Introduction Macros are string substitution for code. It happens during the preprocessing phase. It is generally not encouraged in c++ nowadays. However, there are still applications where it coul...

C++ - Lvalue and Rvalue Reference

Reference Lifetime

Reference Lifetime const lvalue reference Simply Extends the Lifetime of A Temp. No copy constructor is called when the temp object is passed in const lvalue reference as a class attribute DO...

C++ - Operators

Basic Operators

Basic Operators a == b == c doesn’t do what you want. Do (a == b && b == c) instead. It’s equivalent to: 1 2 3 int a = 5, b = 5, c = 5; bool res = a == b; res = res == c;

C++ - Bit Operations

A Mumbo Jumbo List About Bit and Byte Operations

Basic Bitwise operations Setting a specific bit to 1. 1 2 uchar desc_byte = 0; desc_byte |= (1 << NUM);

C++ - Erase Remove

An efficient in-place way to remove elements in a container

Introduction It’s a common task to remove items in a sequential container that fits a criteria, like an array, vector, queue, etc. For this scenario in C++, it’s advised to use the erase-remove id...

C++ - Datatypes

std::optional, structural binding, iostream, namespace, union, std::variant, std::type_index, std::any, ptrdiff_t

POD Types char or unsigned char is 1 byte float, int, are usually 4 bytes (32 bits) on a 64 bit machine double is 8 bytes (64 bits) long double is 8, or even 16 bytes Common Bugs Ob...

Deep Learning - Designing Machine Learning Systems Notes

Machine learning is to learn complex patterns from existing data, and use them to predict on unseen data. MLOps is a set of tools and best practices for bringing ML into production. ML Algorithms a...

Deep Learning - Mask RCNN Hands-On

TODO

Motivation Matterport’s 2018 Mask R-CNN implementation is able to do: real-time instance segmentation, object classification and human body keypoint detection. A bit of hi...

Deep Learning - MobilenetV2 Hands-On

Inverted Skip Connection, Multiclass Classification

MobileNet Architectures In MobileNet V1, the architecture is quite simple: it’s first a conv layer, then followed by 13 sets of depthwise-separable layers. Finally, it has an avg pool, FC, and a s...

Deep Learning - Deeplab Series Theories

Deeplab, ASPP

Motivation UNet has restricted receptive fields. It is sufficient for identifying local areas such as tumors (in medical images). When we learn larger image patches, UNet was not sufficient. Deep...