Public

FLIM — Public Tier

(5 scenes)

Full-access development tier with all data visible.

Measurements (y), ideal forward operator (H), spec ranges, ground truth (x_true), and true mismatch spec.

Load HDF5 → compare reconstruction vs x_true → check consistency → iterate.

Reconstructed signals (x_hat) and corrected spec as HDF5.

✓

True spec visible — use these exact values for Scenario III oracle reconstruction.

Parameter	Spec Range	True Value	Unit
irf_width	-20.0 – 40.0	10.0	ps
time_bin	-5.0 – 10.0	2.5	ps
afterpulsing	-0.005 – 0.01	0.0025

Method: CPU_baseline — Mismatch parameter: nominal

13.27 dB

SSIM 0.3991

11.88 dB

SSIM 0.2385

19.75 dB

SSIM 0.4579

Per-scene breakdown (4 scenes)

Scene	PSNR I	SSIM I	PSNR II	SSIM II	PSNR III	SSIM III
scene_00	17.02	0.3838	14.94	0.2440	20.04	0.5013
scene_01	14.23	0.2969	12.92	0.2098	19.23	0.5480
scene_02	8.52	0.3871	8.01	0.2308	20.11	0.3417
scene_03	13.30	0.5284	11.66	0.2694	19.61	0.4408

#	Method	Score	PSNR	SSIM	Consistency	Trust	Source
1	DiffFLIM + gradient	0.842	37.05	0.978	0.9	✓ Certified	Gao et al., NeurIPS 2024
2	PhysFLIM + gradient	0.825	36.17	0.974	0.87	✓ Certified	Chen et al., Nat. Photonics 2024
3	SwinFLIM + gradient	0.807	34.04	0.961	0.91	✓ Certified	Zhang et al., Biomed. Opt. Express 2023
4	TransFLIM + gradient	0.787	32.53	0.948	0.91	✓ Certified	Wang et al., Nat. Methods 2022
5	FLIMJ + gradient	0.758	30.97	0.93	0.88	✓ Certified	Li et al., Nat. Methods 2022
6	DnCNN-FLIM + gradient	0.718	28.68	0.894	0.87	✓ Certified	Smith et al., Nat. Methods 2019
7	RLD-FLIM + gradient	0.686	26.14	0.836	0.95	✓ Certified	Ballew & Demas, Anal. Chem. 1989
8	MLE-FLIM + gradient	0.600	22.82	0.724	0.91	✓ Certified	Grinvald & Steinberg, Anal. Biochem. 1974
9	Phasor-FLIM + gradient	0.575	21.66	0.675	0.94	✓ Certified	Digman et al., Biophys. J. 2008

y H_ideal spec_ranges x_true true_spec

Format: HDF5

Scenes: 5

0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖)

PSNR: 40% SSIM: 40% Consistency: 20%