hanyuyuan/STSF-TPLS-weights

STSF+TPLS — pretrained checkpoints

Pretrained weights for the paper "The Lift Spectrum: How Measurement-to-Space Adaptivity Shapes Robustness in Image-Free Single-Pixel Sensing" (arXiv:2607.22077).

STSF (spatiotemporal soft-fusion network) performs image-free segmentation directly from 1D single-pixel measurement sequences (512 coded measurements at a 3.13 % sampling rate for 128×128 scenes), trained under TPLS (task-prioritized loss scheduling). The checkpoints perform content-adaptive direct lifting for mask prediction: TPLS's learned reconstruction branch supplies scheduled supervision whose loss shapes the shared lifted representation during training. Its reconstructed output is not consumed by the segmentation head. Released checkpoints retain the branch, and the current code computes and returns aux_recon; it can be pruned for deployment. This repository holds the main-model checkpoints only — the released STSF+TPLS arms. Training/evaluation code, configs, and all baselines live in the code repository: https://github.com/Hanyuyuan6/STSF-TPLS.

Checkpoints

All files are inference-only checkpoints/<run>/best.pth exports (about 147.6 MB each; 1.48 GB total). The full training config is embedded in each checkpoint under config, but optimizer and scheduler state are intentionally omitted, so these files cannot resume training.

File	Dataset / data protocol	Seed	Role
checkpoints/rev_carvana_tpls_s42/best.pth	Carvana	42	simulation
checkpoints/rev_carvana_tpls_m512_s43/best.pth	Carvana	43	simulation
checkpoints/rev_carvana_tpls_m512_s44/best.pth	Carvana	44	simulation
checkpoints/rev_mnist_tpls_s42/best.pth	MNIST	42	simulation
checkpoints/rev_mnist_tpls_m512_s43/best.pth	MNIST	43	simulation
checkpoints/rev_mnist_tpls_m512_s44/best.pth	MNIST	44	simulation
checkpoints/rev_wbc_tpls_s42/best.pth	WBC — paper-legacy-v1 (231/58/60)	42	simulation
checkpoints/rev_wbc_tpls_s43/best.pth	WBC — paper-legacy-v1 (231/58/60)	43	simulation
checkpoints/rev_wbc_tpls_s44/best.pth	WBC — paper-legacy-v1 (231/58/60)	44	simulation
checkpoints/rev_mnist_tpls_m512_s42/best.pth	MNIST	42	optical bench

Exact byte counts and Git LFS SHA-256 object IDs are in MANIFEST.json. The legacy MD5 values from the original release metadata are retained there for compatibility; they were not recomputed during this documentation update.

Usage

Clone the code repository, then point its evaluation entry at a downloaded checkpoint — the loading path, model construction (GRUUNetPP), and metrics are all handled by scripts/evaluate.py there; the checkpoint's embedded config reproduces the exact architecture. Example download:

from huggingface_hub import hf_hub_download
ckpt_path = hf_hub_download(
    repo_id="hanyuyuan/STSF-TPLS-weights",
    filename="checkpoints/rev_carvana_tpls_s42/best.pth",
    revision="22a6e8ee71212ed4574b1a35a5c27e0681219dba",
)

Load release checkpoints in fail-closed mode:

import torch

ckpt = torch.load(ckpt_path, map_location="cpu", weights_only=True)

If weights_only=True fails, stop and verify the immutable commit, SHA-256, code revision, and PyTorch environment. Do not automatically retry the same file with weights_only=False; that enables general pickle execution. No release tag exists in either repository; the example and MANIFEST.json therefore use the real immutable Hugging Face commit 22a6e8ee71212ed4574b1a35a5c27e0681219dba and the audited code commit 3552378c260add875790850505a049867f7c4990.

Operating envelope

These checkpoints reproduce and extend the experiments of arXiv:2607.22077 using the linked repository's evaluation code and the checkpoint's embedded config. They run under a fixed acquisition:

• Acquisition: natural-order Sylvester Hadamard sampling, 128x128 scenes, M = 512 measurements (3.13 % sampling). Input construction and normalization must match the linked code and embedded config; other operators, mask orderings, resolutions, or measurement counts require new validation.
• Exports: inference weights only — optimizer and scheduler state were stripped, so these files cannot resume training.

Not for clinical use. The WBC checkpoint is a research artifact, not a diagnostic or clinical decision-making system. None of these checkpoints is intended for safety-critical deployment.

Scope and provenance

• Weights only. Datasets are not redistributed here; the code repository's README documents how to obtain Carvana, MNIST, and the WBC dataset and how splits are built.
• WBC protocol. All three WBC checkpoints use paper-legacy-v1, the deterministic 231/58/60 released split. The code repository also provides a collision-safe 400-sample full-v2 mode for future experiments; the released checkpoints are not linked to that different protocol.
• The measurement model is fixed natural-order (Sylvester) Hadamard sampling at M = 512 (3.13 % for 128²); these checkpoints assume that acquisition.
• Code license: see the code repository. Weights license: CC-BY-4.0 (attribution: cite the paper below).

Citation

Preprint: https://arxiv.org/abs/2607.22077

@article{han2026liftspectrum,
  title         = {The Lift Spectrum: How Measurement-to-Space Adaptivity Shapes Robustness
                   in Image-Free Single-Pixel Sensing},
  author        = {Han, Yuyuan and Li, Jingwei and Zhang, Xiaoxia and Qiu, Long and Wang, Chong and Hao, Wenxuan
                   and Han, Jiangyu and Yao, Xinyu and He, Yuchen and Chen, Hui and Liu, Jianbin
                   and Zheng, Huaibin},
  journal       = {arXiv preprint arXiv:2607.22077},
  year          = {2026},
  eprint        = {2607.22077},
  archivePrefix = {arXiv},
  primaryClass  = {eess.IV},
}