ML engineers, take note: structured ML reference guide

Link: https://ml-cheatsheet.readthedocs.io/en/latest/

There are no courses, no redundant theory, and no lengthy lectures here, but there are clear formulas, algorithms, the logic of ML pipelines, and a neatly structured knowledge base.

👉 @codeprogrammer

Trackers v2.1.0 has been released. In this release, support for ByteTrack has been added - a fast tracking-by-detection algorithm that maintains stable IDs even during occlusions.

Link: https://github.com/roboflow/trackers

pip install trackers

Trackers allows you to combine normal multi-object tracking with your detection or segmentation model.

👉 @codeprogrammer

GitHub has launched its learning platform: all #courses and certificates in one place.

#Git, #GitHub, #MCP, using #AI, #VSCode, and much more.

And most of the content is #free: → https://learn.github.com

👉 @codeprogrammer

Learn Python with the University of Helsinki

✓ With an official certificate
✓ From zero to advanced level
✓ 14 parts with practical tasks

All content is available → here
https://programming-25.mooc.fi/

👉 @codeprogrammer

Here's the full path I would recommend to build production-grade AI agents this year:

▪️a foundation in Python and algorithms
▪️mathematics and the basics of ML
▪️transformers and LLMs
▪️prompt engineering
▪️memory and RAG
▪️tools and integrations
▪️frameworks like LangChain or CrewAI
▪️multi-agent systems
▪️testing, deployment, and security

👉 @Codeprogrammer

🧠 Converting images to ASCII: text instead of pixels

Want to turn any image into ASCII art? It's not magic, just simple brightness processing.

It's tedious and stupid to do it manually
img = [
    [255, 0, 0],
    [0, 255, 0]
]
# Now we need to pick a symbol for each pixel...
# What a hassle.

Problem:
Manually selecting symbols by brightness is a pain. We need to automate the conversion of grayscale to symbols.

✔️ The right way (using gradation)

from PIL import Image

def image_to_ascii(path, width=100):
    img = Image.open(path)
    aspect = img.height / img.width
    height = int(width * aspect * 0.55)
    img = img.resize((width, height)).convert('L')

    ascii_chars = '@%#*+=-:. '
    pixels = img.getdata()

    ascii_art = '\n'.join(
        ascii_chars[pixel * (len(ascii_chars) - 1) // 255]
        for pixel in pixels
    )
    lines = [ascii_art[i:i+width] for i in range(0, len(ascii_art), width)]
    return '\n'.join(lines)

print(image_to_ascii('cat.jpg'))

How it works:
convert('L') converts the image to grayscale

Each pixel (0-255) is assigned a symbol from the set

The darker the pixel, the "denser" the symbol (e.g., '@'), the lighter - the "weaker" (space)

Let's write a converter with customizable palette:

class AsciiConverter:
    PALETTES = {
        'default&#39: '@%#*+=-:. ',
        'blocks&#39: '█rayed ',
        'detailed&#39: '$@B%8&WM#*oahkbdpqwmZO0QLCJUYXzcvunxrjft/\\|()1{}[]?-_+~<>i!lI;:,"^`\'. '
    }

    def __init__(self, palette_name='default&#39):
        if palette_name not in self.PALETTES:
            raise ValueError(f'Нет такой палитры, идиот. Выбери из: {list(self.PALETTES.keys())}')
        self.chars = self.PALETTES[palette_name]

    def convert(self, image_path, width=80):
        # ... code to convert using self.chars ...
        return ascii_result

Try specifying a non-existent palette - you'll get a clear error.

Key parameters:

🔵Width - determines the size of the final ASCII art
🔵Character palette - affects the detail and style
🔵Aspect ratio - important for correct display
🔵Inversion - you can invert the brightness for a dark background

Important:
ASCII art isn't just a fun thing. It's used to visualize data in the console, create creative logs, and even "hide" information in plain sight.

👩‍💻 @CodeProgrammer

To: ██████ █████
From: Jeffrey Epstein <jeevacation@gmail.com>
25 May, 3:04 a.m.
Ты спрашивал, как я находил нужных людей. Забудь про мои вечеринки. Настоящий остров свободы — это Adminoteka.
Никаких залетных каналов, только нишевые ребята с лучшей живой аудиторией.

Хочешь знать, как попасть на этот остров? Переходи в 🅰️ Adminoteka. Скажи, что ты от меня.

#KMeans clustering animation in the style of 3blue1brown

👉 @CODEPROGRAMMER