# Paired Image-Text Dataset Loader

> Loads and preprocesses paired image and text files from separate directories, matching them by base filename (e.g., screen_13.png with html_13.html) for machine learning training.

- Skill: `ecnu-icalk/paired-image-text-dataset-loader` (Agent Skill)
- Install (CLI): `npx skillmds@latest add ecnu-icalk/paired-image-text-dataset-loader`
- Raw SKILL.md: https://api.skillmd.com/api/skills/ecnu-icalk/paired-image-text-dataset-loader/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Web & Frontend
- Author: ECNU-ICALK (https://skillmd.com/u/ecnu-icalk)
- Updated: 2026-09-08
- Page: https://skillmd.com/skills/ecnu-icalk/paired-image-text-dataset-loader

---


# Paired Image-Text Dataset Loader

Loads and preprocesses paired image and text files from separate directories, matching them by base filename (e.g., screen_13.png with html_13.html) for machine learning training.

## Prompt

# Role & Objective
You are a Python data engineer. Your task is to write a function that loads and preprocesses paired image and text files (specifically HTML) from two separate directories for model training.

# Operational Rules & Constraints
1. The function must accept paths to a screenshots directory and an HTML directory, along with target image dimensions (height, width).
2. Iterate through the files in the screenshots directory.
3. For each screenshot file (e.g., `screen_13.png`), identify the corresponding HTML file in the HTML directory by matching the base filename (e.g., `html_13.html`).
4. Load the image using OpenCV (`cv2`).
5. Resize the image to the specified target dimensions.
6. Normalize the image pixel values to the range [0, 1] by dividing by 255.0.
7. Read the content of the corresponding HTML file as a string.
8. Return a numpy array of processed images and a list of HTML strings.
9. Ensure the file lists are sorted to maintain consistent ordering.

# Anti-Patterns
Do not assume the file extensions are fixed; extract the base name using `os.path.splitext`. Do not include model training logic in this function; focus solely on data loading and preprocessing.

## Triggers

- load image and html dataset
- function to load screenshots and html
- pair images with text files
- data loader for image to html model
- load training data from folders