Dev

Mammography — 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	CT-ViT + gradient	0.805	35.0	0.968	0.84	✓ Certified	Guo et al., NeurIPS 2024
2	CTFormer + gradient	0.799	34.5	0.964	0.84	✓ Certified	Li et al., ICCV 2024
3	DiffusionCT + gradient	0.779	33.33	0.955	0.82	✓ Certified	Kazemi et al., ECCV 2024
4	Score-CT + gradient	0.772	32.51	0.948	0.84	✓ Certified	Song et al., NeurIPS 2024
5	DOLCE + gradient	0.764	31.91	0.942	0.84	✓ Certified	Liu et al., ICCV 2023
6	DuDoTrans + gradient	0.700	27.35	0.866	0.9	✓ Certified	Wang et al., MLMIR 2022
7	PnP-ADMM + gradient	0.690	27.47	0.869	0.84	✓ Certified	Venkatakrishnan et al., IEEE GlobalSIP 2013
8	RED-CNN + gradient	0.687	27.6	0.872	0.81	✓ Certified	Chen et al., IEEE TMI 2017
9	Learned Primal-Dual + gradient	0.683	27.52	0.87	0.8	✓ Certified	Adler & Oktem, IEEE TMI 2018
10	FBPConvNet + gradient	0.678	27.35	0.866	0.79	✓ Certified	Jin et al., IEEE TIP 2017
11	PnP-DnCNN + gradient	0.670	26.76	0.852	0.81	✓ Certified	Zhang et al., IEEE TIP 2017
12	TV-ADMM + gradient	0.646	25.85	0.828	0.78	✓ Certified	Sidky et al., Phys. Med. Biol. 2008
13	FBP + gradient	0.622	24.45	0.784	0.82	✓ Certified	Kak & Slaney, IEEE Press 1988

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%