---
license: mit
pretty_name: DLAA Trainer Image Dataset
task_categories:
- image-to-image
- image-classification
language:
- en
tags:
- image-restoration
- anti-aliasing
- dlaa
- texture
- computer-vision
- comfyui
- solricks
size_categories:
- 1K<n<10K
---

# DLAA Trainer Image Dataset

This dataset contains a cleaned collection of high-resolution portrait, nature, object, texture, and geometric images prepared for training and evaluating lightweight image restoration, anti-aliasing, and texture refinement models.

It was prepared as part of the `DLAANet.safetensors` and `DLAATexture.safetensors` experimentation workflow, especially for improving edge quality, fine detail stability, texture preservation, geometric consistency, and visual cleanup in image and video pipelines.

## Dataset Structure

The dataset is stored under:

```text
dataset/
├── image_001.jpg
├── image_002.jpg
├── image_003.png
└── ...