harleyszhang/cv_note: a computer vision engineer's Chinese study notes, now winding down

What cv_note is, and the reader it was written for

cv_note is not a library. It is a Markdown knowledge base in Chinese, aimed at people preparing for or working in computer vision algorithm roles, especially in the Chinese job market. The README describes its origin plainly: it began as an internship referral sheet and a list of companies hiring for campus recruitment, then turned into a personal record of the technical stack a CV algorithm engineer is expected to know, plus some interview notes.

The scope is wider than the name suggests. The top-level directories run from 1-computer_basics and 2-programming_language through 3-machine_learning, 4-deep_learning, 5-computer_vision, 6-model_compression, 7-high-performance_computing, 8-model_deploy, and end at interview_summary. A separate file, cv算法工程师成长路线.md, lays out a learning route, and the README lists a study order that starts with one complete foreign course, one machine learning or deep learning textbook, one programming language, and one framework.

The intended reader is someone who already knows they want to work on detection, segmentation, quantization or deployment, and wants a map of the territory rather than a tutorial series. If you are looking for a Python package to install, this is the wrong repository, and the README never pretends otherwise.

How the notes are organised, and why that structure matters

The repository is a directory tree of Markdown files. Numbered folders encode a suggested sequence, and SUMMARY.md plus book.json indicate the content was also meant to be rendered as a book, most likely through GitBook. The README notes that GitHub renders LaTeX formulas directly, and recommends the MathJax Plugin for Github, Typora, or VS Code with the Markdown+Math extension when formulas do not display fully.

That is the whole mechanism: files on disk, rendered by whatever Markdown tool you prefer. There is no build step that produces a site, no search index, and no generated API documentation. The homepage field is empty, and the README links to a personal blog at armcvai.com rather than to a project site.

The practical consequence is that navigation depends on the reader. A numbered folder name tells you the intended order, but nothing tells you which files are complete. The README states that some knowledge points are not finished, without saying which. Treat the tree as a syllabus and verify each file before you plan study time around it.

Installing and reading cv_note locally

There is nothing to install in the software sense. The README gives no package, no pip command, and no release artifact; the repository is the deliverable. The steps below clone it and render the Markdown with a tool the README itself recommends.

Clone the repository first. This is the only command needed to obtain the content.

git clone https://github.com/HarleysZhang/cv_note.git
cd cv_note

The README suggests Typora for local reading, and VS Code with the Markdown+Math extension as an alternative when LaTeX formulas do not render correctly. Open the folder in either tool and start from the numbered directories.

code cv_note

What you should see is a tree of Chinese-language Markdown files with the directory names listed above. There is no server to start and no configuration file to edit. If you prefer reading on GitHub, the README notes that formulas may need the MathJax Plugin for Github in Chrome, and that downloading the repository locally is the fallback.

The README says the project is being abandoned

This is the first thing to know before spending time here. The README states directly that the project is gradually being abandoned and that most content is no longer updated. It redirects readers interested in deep learning, large model inference, and inference framework development to two other repositories, dl_note and lite_llama.

The company list has already been hollowed out for the same reason. The README explains that the spring recruitment internship table written in 2019 had mostly expired and that the author no longer had the energy to maintain it, so it was removed from the repository homepage. What remains is a static table of employer names sorted into top-tier companies, internet companies, AI unicorns, and other large firms. It has no dates, no links to postings, and no indication of which entries are still hiring.

The last push to the repository was on 2026-05-10, which is more than six months before today. The repository is not archived, but the README's own words and the push date point the same way. Anyone treating cv_note as a current reference for a job search is working from material the author has already declared stale.

The paid Triton course sitting at the top of the README

The first section of the README is not about the notes at all. It advertises a course on building a large model inference framework, priced at 499, developed with the author of a separate deep learning inference framework course. The README lists the framework's claimed features: Triton kernels written in a PyTorch-like syntax, PyTorch-based memory management, support for FlashAttention V1, V2 and V3, GQA, PageAttention, fused operators such as KV linear layer fusion, and adaptation to llama, qwen2.5 and llava1.5 models. It also claims up to a 4x speedup over the transformers library on llama3 1B and 3B models.

Those numbers come from the README's marketing text, not from an independent measurement, and the course is a commercial product bought by scanning a WeChat QR code. Nothing in the repository tree suggests the course code lives here; the top-level entries are notes, a licence, a book configuration, and data assets.

For a reader evaluating cv_note as an open source project, the practical point is that the most prominent content on the page is an advertisement. That does not make the notes worthless, but it does mean the README is optimised for selling the course rather than for documenting the repository.

How cv_note compares with a structured curriculum

The obvious alternative is a formal course with assignments and a grader, which is what the README itself recommends as the starting point: complete one foreign course from videos through homework to a project, read one textbook end to end, and only then use notes like these. cv_note is the opposite kind of artifact. It is a personal index, written in Chinese, with uneven depth across topics and no exercises.

A second alternative is the author's own successor repositories. The README directs readers to dl_note for deep learning and large model inference, and to lite_llama for inference framework development. If your interest is model deployment and high performance computing, those are the maintained destinations; cv_note is the older, broader, and now largely frozen notebook.

The difference in approach is worth stating plainly. A course imposes an order and checks your understanding. A notes repository assumes you already know what you are looking for and only need pointers. cv_note fits the second case, and only in Chinese.

Licence, maintenance cost and what to verify

The repository carries Apache-2.0, and the README renders an Apache-2.0 badge. That is a permissive licence for the text and any code snippets in the notes, but it says nothing about the third-party figures, screenshots or quoted passages the README acknowledges were drawn from GitHub projects, blogs and books. If you intend to reuse a diagram or a long passage, check its origin yourself; the licence file covers the repository, not necessarily every asset inside it. This is a description of what the licence states, not legal advice.

The maintenance cost for a user is low in one sense and high in another. There is nothing to upgrade, no dependency to pin, no release to track. On the other hand, the content will not be corrected. The README admits unfinished sections, the company table is undated, and the author has said most content is no longer updated.

Before relying on a chapter, open the specific file and confirm it is written rather than stubbed. If your topic is deep learning or inference frameworks, check whether the successor repositories already cover it better. And if you need English material or a runnable project, look elsewhere from the start.

Editorial conclusion

Adopt cv_note only as a reading list and directory index, not as a maintained reference: the README states the project is gradually being abandoned and that most content is no longer updated, and the last push was on 2026-05-10. Anyone who needs current material on deep learning or large model inference should go to the successor repositories the README names, dl_note and lite_llama. Before relying on any chapter, open the directory and check whether the topic you need is actually written, because the README admits some knowledge points are unfinished. Skip it if you want English-language material, a single installable package, or a course you can run; the paid Triton and PyTorch inference framework course is a separate product sold through a WeChat QR code, not part of this repository.

Frequently asked questions

Official sources

1. harleyszhang/cv_note on GitHub
2. Issues
3. License: Apache-2.0
4. README