Alp3r3n

Alp3r3n / cuda-dataset-cleaner

Public

CUDA/C++ tool for GPU-accelerated image dataset quality scanning and CPU vs GPU benchmarking.

10
0
100% credibility
Found May 11, 2026 at 10 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
C++
AI Summary

A tool that scans image folders to assess quality metrics like brightness, contrast, sharpness, and saturation, then generates reports recommending images to keep, review, or delete.

How It Works

1
🔍 Discover the image checker

You find this helpful tool while searching for a quick way to spot bad photos in large collections for your projects.

2
💻 Prepare the tool

You follow simple steps to set up the image quality checker on your computer.

3
📁 Choose your photo folder

You select the folder full of images you want to check.

4
⚡ Start the fast scan

You launch the check and see it zip through your pictures, measuring brightness, sharpness, contrast, and more.

5
📊 See the overview

A summary appears showing how many images look good, need a look, or should go.

✅ Review your reports

You open the detailed lists to sort images easily, keeping only the best for your work.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 10 to 10 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is cuda-dataset-cleaner?

This CUDA/C++ CLI tool scans image datasets (JPG, PNG, BMP) for quality issues like blur, low contrast, over/underexposure using GPU-accelerated metrics alongside CPU baselines via OpenCV. It flags problems conservatively, assigns 0-100 quality scores, and outputs keep/review/delete recommendations in JSON and CSV reports, plus aggregate CPU vs GPU benchmarking. Built for fast triage of noisy datasets that hurt ML/CV training, it processes folders recursively with configurable thresholds.

Why is it gaining traction?

It delivers tangible GPU speedups—up to 6x on COCO's 5k images—while doubling as a CUDA C++ GitHub example for real-world image processing and benchmarking. Unlike generic cleaners, its conservative logic routes edge cases to human review, avoiding over-deletion, and includes per-image timings for easy CPU/GPU comparisons. Developers grab it for the cuda c++ toolkit showcase in the CUDA C++ developer's toolbox.

Who should use this?

CV/ML engineers prepping large image datasets for training, needing a quick gpu-accelerated scanner before manual cleanup. CUDA/C++ devs benchmarking custom kernels against OpenCV on image quality metrics. Teams evaluating CUDA/C++ GitHub projects for dataset tools or gpu vs cpu performance baselines.

Verdict

Grab it for prototyping dataset quality scanning or as a CUDA C++ programming guide GitHub benchmark—docs and recent optimizations impress despite 10 stars and 1.0% credibility score. Too early for heavy production; fork and batch it up for scale.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.