README.md

metadata

license: cc
language:
  - en
tags:
  - INR
  - 2d
  - 3d
  - image
  - voxel
size_categories:
  - 1M<n<10M
configs:
  - config_name: div2k
    data_files:
      - split: '0'
        path: DIV2K/0064.png
      - split: '1'
        path: DIV2K/0007.png
      - split: '2'
        path: DIV2K/0010.png
      - split: '3'
        path: DIV2K/0029.png
      - split: '4'
        path: DIV2K/0063.png
      - split: '5'
        path: DIV2K/0072.png
      - split: '6'
        path: DIV2K/0079.png
      - split: '7'
        path: DIV2K/0088.png
      - split: '8'
        path: DIV2K/0093.png
      - split: '9'
        path: DIV2K/0131.png
  - config_name: ct
    data_files:
      - split: '1234'
        path: chest.png
  - config_name: spheres
    data_files:
      - split: '1234'
        path: SparseSphereSignal/1234/*.npy
      - split: '2024'
        path: SparseSphereSignal/2024/*.npy
      - split: '5678'
        path: SparseSphereSignal/5678/*.npy
      - split: '7618'
        path: SparseSphereSignal/7618/*.npy
      - split: '7890'
        path: SparseSphereSignal/7890/*.npy
  - config_name: bandlimited
    data_files:
      - split: '1234'
        path: BandlimitedSignal/1234/*.npy
      - split: '2024'
        path: BandlimitedSignal/2024/*.npy
      - split: '5678'
        path: BandlimitedSignal/5678/*.npy
      - split: '7618'
        path: BandlimitedSignal/7618/*.npy
      - split: '7890'
        path: BandlimitedSignal/7890/*.npy
  - config_name: sierpinski
    data_files:
      - split: '0.1'
        path: sierpinski_triangle/*0.npy
      - split: '0.2'
        path: sierpinski_triangle/*1.npy
      - split: '0.3'
        path: sierpinski_triangle/*2.npy
      - split: '0.4'
        path: sierpinski_triangle/*3.npy
      - split: '0.5'
        path: sierpinski_triangle/*4.npy
      - split: '0.6'
        path: sierpinski_triangle/*5.npy
      - split: '0.7'
        path: sierpinski_triangle/*6.npy
      - split: '0.8'
        path: sierpinski_triangle/*7.npy
      - split: '0.9'
        path: sierpinski_triangle/*8.npy
  - config_name: star_target
    data_files:
      - split: '1234'
        path: star_resolution_target.npy

Signal Dataset Loader

This repository provides a small collection of synthetic and real signals—both 2D and 3D—used for compression, reconstruction.

---

Quick Start

All classes share the call signature

(dimension, length, bandlimit, seed, generate=True, super_resolution=False, sparse=False)

• dimension – 2 or 3 (ignored when not applicable)
• length – 1000
• bandlimit – fractional control variable (0.1 – 0.9 in 0.1 increments for BandlimitedSignal, SparseSphereSignal, and Sierpinski; interpretation varies per class)
• seed – ensures deterministic generation and consistent file paths (for BandlimitedSignal and SparseSphereSignal the repository ships five predefined seeds: 1234, 2024, 5678, 7890, 7618)
• generate – True = create new signal, False = load cached .npy
• super_resolution / sparse – optional toggles (see catalog below)

---

Signal Catalog

Synthetic Signals (~1 M values each)

Class	Dim	Description
BandlimitedSignal	2D / 3D	Uniform noise passed through a circular low‑pass filter; nine preset cut‑offs yield progressively higher spatial frequencies
SparseSphereSignal	2D / 3D	Random circles/spheres occupying a fixed volume fraction; sphere radius inversely proportional to bandlimit
Sierpinski	2D	Classic Sierpinski triangle rendered at depths 0 – 9, depth = int(bandlimit*10)−1
StarTarget	2D	Star‑shaped resolution target with alternating wedges; default 40 solid wedges (80 spokes total)

Real‑World Signals

Class	Notes
RealImage	Ten DIV2K images (DIV2K/00xx{,x4}.png). super_resolution=False loads the bicubic ×4 LR image; True loads the HR counterpart
Voxel_Fitting	Stanford Dragon voxel grid. sparse=True keeps only surface voxels; False loads full occupancy. super_resolution picks a higher‑res scan
CTImage	Single axial chest CT slice (chest.png), loaded as grayscale float32

---

Adding Your Own Signal

1. Subclass the same pattern and expose a self.signal NumPy array.
2. Save deterministic outputs to <ClassName>/<seed>/ so they can be re‑loaded with generate=False.
3. Keep the in‑memory footprint under ~1 M elements for apples‑to‑apples comparisons.

---

Citation & Licensing

If you use this loader in academic or industrial work, please cite:

@article{kim2025grids,
  title   = {Grids Often Outperform Implicit Neural Representations},
  author  = {Kim, Namhoon and Fridovich-Keil, Sara},
  journal = {arXiv preprint arXiv:2506.11139},
  year    = {2025}
}

Code and synthetic assets are released under the Creative Commons CC‑BY‑4.0 license. Real images remain subject to the terms of their original datasets (e.g., DIV2K).