harleyszhang/cv_note

What is cv_note?

cv_note is a personal knowledge base and career guide for Computer Vision (CV) algorithm engineers — the people who build systems that let computers "see" and understand images. It was created by a developer who goes by HarleysZhang.

The project started as a practical job-hunting toolkit: a list of companies that hire CV algorithm engineers, plus a collection of interview questions and the author's own interview experiences. Over time, it grew into a broader set of technical study notes covering the skills a CV engineer needs.

What's inside

The repository is organized as a learning path with folders for:

• Computer fundamentals (operating systems, networks, etc.)
• Programming languages (with a Python 3 summary)
• Data structures and algorithms
• Machine learning and deep learning (including a deep learning basics summary)
• Computer vision (image recognition, object detection, semantic segmentation)
• Model compression and quantization (making AI models smaller and faster)
• High-performance computing
• Model deployment (getting models to run in production)
• Interview summaries for image algorithm roles

There's also a dedicated learning roadmap file and a mind map of the knowledge points a CV engineer should master.

The README includes a large table of companies that hire for these roles, grouped by tier: top tech companies (Baidu, Alibaba, Tencent, Microsoft, Google), internet companies (Meituan, ByteDance, JD.com), AI startups (SenseTime, Horizon Robotics, Momenta), and other large firms (Hikvision, DJI, Huawei, etc.).

The paid course (a major focus of the README)

The README devotes significant space to promoting a paid course (499 yuan) that the author co-created with the author of another project called "Self-Made Deep Learning Inference Framework." The course teaches you to build a large language model (LLM) inference framework from scratch — the software that runs AI models like ChatGPT after they've been trained.

Key claims about the course:

• Built with Triton + PyTorch — Triton is a programming language that makes GPU acceleration easier without writing complex CUDA code
• Covers the full offline inference pipeline for large models
• Implements FlashAttention V1/V2/V3 (a fast attention algorithm), GQA (a memory-saving attention variant), and PageAttention (a memory management technique)
• Supports recent models like Llama, Qwen2.5, and Llava1.5
• Claims up to 4x speedup on Llama 3 1B and 3B models compared to the standard transformers library
• Includes categorized interview questions for high-performance computing / inference framework roles (2025 edition)

Contact information (a WeChat QR code) is provided for purchasing.

Current status: being deprecated

The README states clearly that this project is being gradually retired. Most content is no longer updated. The author directs readers to two other repositories for ongoing work:

• dl_note — for deep learning and LLM inference knowledge
• lite_llama — for LLM inference framework development

Other details

• The author maintains a WeChat public account called "嵌入式视觉" (Embedded Vision) and has a personal blog at armcvai.com, plus profiles on Zhihu, CSDN, CNBlogs, and Juejin.
• The README suggests that if GitHub doesn't render LaTeX formulas properly, you can use a browser plugin or download the repo and read it with Typora or VSCode.
• The project is licensed under Apache 2.0.

In short

cv_note is a Chinese-language career and study guide for aspiring CV algorithm engineers — part job-hunting reference, part technical notebook. It's no longer actively maintained, but it still contains a structured learning path and a company list. The README's main current purpose seems to be directing readers to the author's newer projects and promoting the paid LLM inference framework course.