Aydin Example Datasets: Microscopy and Natural Images for Image Denoising Benchmarks

A curated collection of 29 example images used by Aydin, a self-supervised image denoising toolkit for n-dimensional microscopy and natural images. These datasets serve as benchmarks for testing and demonstrating Aydin's classical, feature-based, and deep-learning denoising algorithms. The collection spans 2D natural images, noisy test images, 3D fluorescence microscopy volumes, 2D+t and 3D+t time-lapse sequences from light-sheet, spinning-disk confocal, and widefield fluorescence microscopy. How to use these files Individual file access (no authentication required): GET https://zenodo.org/api/records/{record_id}/files/{filename}/content From Python (with Aydin installed): from aydin.io.datasets import newyork; image = newyork() Large TIFF files have been losslessly compressed using TIFF zstd compression with horizontal differencing predictor. They can be read directly by tifffile, ImageJ/FIJI, and most modern TIFF readers without any decompression step. File inventory 2D Natural Images (monochrome, uint8) newyork.png — 1024×1024, 0.7 MB. Original photograph of New York City by Loic A. Royer. CC BY 4.0. scafoldings.png — 1024×1536, 1.0 MB. Original photograph of scaffoldings by Loic A. Royer. CC BY 4.0. andromeda.png — 1024×1024, 0.5 MB. Isaac Roberts, 1888 long-exposure photograph of M31 (Andromeda Galaxy), taken with a 20-inch reflecting telescope at Maghull, Liverpool. Public domain (copyright expired; photographer died 1904). Wikipedia | Science Museum Group pollen.png — 600×788, 0.3 MB. Scanning electron micrograph of pollen grains by Louisa Howard & Charles Daghlian, Dartmouth Electron Microscope Facility. Public domain (released by Dartmouth College). Source characters.jpg — 1355×875, 0.3 MB. Comparative chart of ancient writing systems (Cuneiform, Egyptian hieroglyphs, Chinese). Public domain (Wikimedia Commons). periodic_noise.png — 473×1024, 0.4 MB. Newspaper scan with moiré noise, from OpenCV documentation. Apache 2.0 / BSD-3. Tutorial 2D Noisy Images fountain.png — 588×402, uint8+alpha, 0.3 MB. Wallace fountain from the University of Edinburgh image processing test library. Public domain / open access. monalisa.png — 568×382, uint8+alpha, 0.2 MB. Leonardo da Vinci — Mona Lisa (c. 1503) with synthetic noise. Public domain. Wikipedia Gauss_noisy.png — 1305×1024, uint8 RGB, 3.2 MB. Portrait of Carl Friedrich Gauss by C.A. Jensen (1840) with synthetic noise. Public domain (artist died 1870). Wikipedia newyork_noisy.tif — 1024×1024, float32, 3.1 MB. Derived from newyork.png with synthetic Poisson + Gaussian noise. CC BY 4.0. Brown_SIDD_chessboard_gray.png — 761×1012, uint16, 1.3 MB. Extracted from the SIDD dataset (Abdelhamed, Lin & Brown, CVPR 2018, Brown University). MIT License. Paper | Download portal 2D RGB Test Image rgbtest.png — 512×512×3, uint8, 12 KB. Synthetic RGB test pattern, project-created. CC BY 4.0. 3D Fluorescence Microscopy Volumes Myers_Tribolium_nGFP_0.1_0.2_0.5_20_13_late.tif — 48×773×665 (ZYX), uint16, 18 MB. Tribolium castaneum embryo expressing nuclear GFP, light-sheet microscopy. From the CARE training data (Myers Lab, MPI-CBG Dresden). CC0 1.0 (public domain). Weigert et al., Nature Methods 2018 SPC0_TM0132_CM0_CM1_CHN00_CHN01.fusedStack.tif — 108×1352×532 (ZYX), uint16, 40 MB. Drosophila melanogaster embryo, dual-view light-sheet (SiMView). Keller Lab, Janelia Research Campus / HHMI. PI-confirmed redistribution. Tomer et al., Nature Methods 2012 Royer_confocal_dragonfly_hcr_drerio_30somite_crop.tif — 75×3×444×331 (ZCYX), uint16, 45 MB. Zebrafish 30-somite embryo, spinning-disk confocal (Dragonfly), HCR fluorescence, 3 channels. Royer Lab, Chan Zuckerberg Biohub. CC BY 4.0. Flybrain_3ch_mediumSize.tif — 294×3×768×768 (ZCYX), uint16, 501 MB. Adult Drosophila brain, 3-channel fluorescence volume. Janelia Research Campus / HHMI. PI-confirmed redistribution. fixed_pattern_noise.tif — 73×400×340 (ZYX), uint16, 4.2 MB. Volume with fixed-pattern noise, acquired on a single-objective light-sheet microscope (soSPIM). Huang Lab (Bo Huang), UCSF. PI-confirmed redistribution. Gao et al., Nature Methods 2015 C2-DrosophilaEggChamber-small.tif — 164×128×128 (ZYX), uint16, 3.5 MB. Drosophila egg chamber (cropped), point-scanning confocal. From the LimeSeg dataset. CC BY 4.0. Machado et al., BMC Bioinformatics 2019 3D Fluorescence — OpenCell / Leonetti Lab From the OpenCell project (Manuel Leonetti Lab, Chan Zuckerberg Biohub) — CRISPR-tagged endogenous proteins imaged in live human cells. CC BY-SA 4.0. Cho et al., Science 2022. Leonetti_p1H8_2_SNCA_PyProcessed_IJClean.tif — 46×2×1024×1024 (ZCYX), uint16, 139 MB. SNCA (α-synuclein). Leonetti_OC-FOV_ARHGAP21_…_stack.tif — 106×600×600 (ZYX), uint16, 54 MB. ARHGAP21. Leonetti_OC-FOV_ANKRD11_…_stack.tif — 106×600×600 (ZYX), uint16, 54 MB. ANKRD11. Leonetti_p4B3_1_TM7SF2_PyProcessed_IJClean.tif — 35×2×1024×1024 (ZCYX), uint16, 106 MB. TM7SF2. Leonetti_p1H10_2_SPTSSA_PyProcessed_IJClean.tif — 36×2×1024×1024 (ZCYX), uint16, 108 MB. SPTSSA. 2D+t Time-Lapse — Cognet Lab (Carbon Nanotube Tracking) Near-infrared fluorescence imaging of single-walled carbon nanotubes (SWCNTs). Cognet Research Group, LP2N, University of Bordeaux / CNRS / IOGS. PI-confirmed redistribution. Danial et al., ACS Nano 2019 | Bhatt et al., Nano Letters 2022 Cognet_1-400fps.tif — 8000×320×320 (TYX), uint16, 797 MB. 400 frames/s, NIR fluorescence. Cognet_1-200fps.tif — 8000×320×320 (TYX), uint16, ~800 MB. 200 frames/s, NIR fluorescence. Cognet_1-100fps.tif — 8000×320×320 (TYX), uint16, 815 MB. 100 frames/s, NIR fluorescence. Cognet_r03-s01-100mW-20ms-175 50xplpeg-173.tif — 20000×80×80 (TYX), uint16, 182 MB. Single nanotube tracking, 20 ms exposure, PEGylated SWCNT. 3D+t and 4D Time-Lapse Maitre_mouse blastocyst_fracking_180124_e3_crop.tif — 10×15×729×729 (TZYX), uint8, 55 MB. Mouse blastocyst, time-lapse light-sheet microscopy. Maitre Lab, Institut Curie. Dumortier et al., Science 2019. CC BY-NC-SA 4.0; PI confirmed redistribution. Not available from any public repository — shared directly by the PI. Hyman_HeLa.tif — 39×47×320×320 (TZYX), uint16, 120 MB. HeLa cell 4D dataset (39 time points, 47 Z-slices). Hyman Lab, MPI-CBG Dresden. PI-confirmed redistribution. Datasets NOT included in this record Downloaded at runtime by Aydin from original source URLs: camera.png — “Cameraman” (MIT, 1978). CC BY-NC. MIT DOME lizard.png — Berkeley BSDS300 #87046. Non-commercial research only. mandrill.tif — USC-SIPI 4.2.03. Copyright unknown. OME-TIFF examples — openmicroscopy.org. Compression Large TIFFs (> 5 MB) losslessly re-compressed with Zstandard (zstd) + horizontal predictor via tifffile. Ratios: 1.3×–4.5×. Pixel data is bit-for-bit identical. Readable by any TIFF reader supporting zstd. Related software Aydin on GitHub Aydin on PyPI Aydin documentation Aydin software DOI