Dev

Multispectral Satellite Imaging — Dev Tier

(3 scenes)

Blind evaluation tier — no ground truth available.

Measurements (y), ideal forward operator (H), and spec ranges only.

Apply your pipeline from the Public tier. Use consistency as self-check.

Reconstructed signals and corrected spec. Scored server-side.

🔒

True spec hidden — estimate parameters from spec ranges below.

#	Method	Score	PSNR	SSIM	Consistency	Trust	Source
1	FlowCompute + gradient	0.743	30.49	0.924	0.84	✓ Certified	Huang et al., ECCV 2025
2	SwinIR + gradient	0.742	30.2	0.92	0.86	✓ Certified	Liang et al., ICCVW 2021
3	Restormer + gradient	0.741	30.51	0.924	0.83	✓ Certified	Zamir et al., CVPR 2022
4	CompFormer + gradient	0.720	29.79	0.914	0.78	✓ Certified	Liu et al., ICCV 2024
5	NAFNet + gradient	0.715	28.36	0.888	0.88	✓ Certified	Chen et al., ICCV 2023
6	DiffusionCompute + gradient	0.714	28.9	0.898	0.83	✓ Certified	Zhang et al., NeurIPS 2024
7	PnP-RED + gradient	0.645	25.37	0.814	0.83	✓ Certified	Romano et al., IEEE TIP 2017
8	Deep Image Prior + gradient	0.644	25.31	0.812	0.83	✓ Certified	Ulyanov et al., CVPR 2018
9	ART + gradient	0.636	25.41	0.815	0.78	✓ Certified	Gordon et al., J. Theor. Biol. 1970
10	LSQR + gradient	0.631	24.58	0.789	0.85	✓ Certified	Paige & Saunders, TOMS 1982
11	PnP-ADMM + gradient	0.607	24.08	0.771	0.79	✓ Certified	Venkatakrishnan et al., 2013
12	Tikhonov + gradient	0.604	23.92	0.766	0.79	✓ Certified	Tikhonov, Doklady Akad. Nauk SSSR 1963
13	Plug-and-Play + gradient	0.524	20.77	0.635	0.81	✓ Certified	Sreehari et al., IEEE TIP 2016

y H_ideal spec_ranges

Format: HDF5

Scenes: 3

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

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