ResMatching

Official implementation of ResMatching, accepted at ISBI 2026.

ResMatching: Noise-Resilient Computational Super-Resolution via Guided Conditional Flow Matching
Anirban Ray, Vera Galinova, Florian Jug [arXiv] | [Interactive Results]

Abstract

Computational Super-Resolution (CSR) in fluorescence microscopy has, despite being an ill-posed problem, a long history. At its very core, CSR is about finding a prior that can be used to extrapolate frequencies in a micrograph that have never been imaged by the image-generating microscope. It stands to reason that, with the advent of better data-driven machine learning techniques, stronger prior can be learned and hence CSR can lead to better results. Here, we present ResMatching, a novel CSR method that uses guided conditional flow matching to learn such improved data-priors. We evaluate ResMatching on 4 diverse biological structures from the BioSR dataset and compare its results against 7 baselines. ResMatching consistently achieves competitive results, demonstrating in all cases the best trade-off between data fidelity and perceptual realism. We observe that CSR using ResMatching is particularly effective in cases where a strong prior is hard to learn, e.g. when the given low-resolution images contain a lot of noise. Additionally, we show that ResMatching can be used to sample from an implicitly learned posterior distribution and that this distribution is calibrated for all tested use-cases, enabling our method to deliver a pixel-wise data-uncertainty term that can guide future users to reject uncertain predictions.

Installation

pip install uv
uv sync

Example Notebook

Start here for the quickest hands-on introduction: notebooks/resmatching_walkthrough.ipynb

This beginner-friendly notebook is aimed at biologists and microscopists and is the easiest way to get a feel for the full ResMatching workflow.

The notebook explains each cell in detail, includes a tiny 2 to 3 epoch training demo, runs small-sample inference with a pre-trained model, computes lightweight metrics, plots calibration, and visualizes the input / ground truth / MMSE / posterior samples.

Datasets

Experiments use the BioSR dataset. The following subsets are supported:

Subset	Structure
ccp	Clathrin-Coated Pits
er	Endoplasmic Reticulum
factin	F-actin
mt	Microtubules
mt_noisy	Microtubules data with additional noise added

Reproducing results

There are two workflows depending on whether you want to train from scratch or use pre-trained checkpoints.

By default, scripts/metrics.py evaluates the current local inference outputs in data/<subset>/<split>_results/. Use --paper-results to download and evaluate the exact saved prediction stacks used for the paper metrics from Zenodo. Full retraining may produce slight variance due to non-deterministic operations.

---

Option A — Train from scratch

Step 1. Download data

# Download all subsets
uv run python scripts/download_data.py

# Or download a specific subset
uv run python scripts/download_data.py --subset ccp

Data is saved to data/<subset>/ by default.

Step 2. Train

uv run python scripts/train.py ccp

The best checkpoint is saved to checkpoints/ccp/best_model.pth. Training runs for 200 epochs by default.

Step 3. Run inference

uv run python scripts/infer.py ccp --checkpoint checkpoints/ccp/best_model.pth

Writes multi-sample TIFFs to data/ccp/test_results/ and data/ccp/val_results/.

Step 4. Compute metrics

# Evaluate the current local inference outputs
uv run python scripts/metrics.py ccp

# Or evaluate the exact paper result stacks from Zenodo
uv run python scripts/metrics.py ccp --paper-results

The default command reads from data/ccp/test_results/. With --paper-results, the script downloads the archive to data/ccp/paper_result_samples/, reads test_result_samples/, and prints PSNR, MicroMS3IM, LPIPS, and FID. Add --split val to evaluate the validation stacks. The Zenodo archive for mt_noisy is named mtNoisy, which the script handles automatically.

Step 5. (Optional) Calibration

uv run python scripts/calibrate.py ccp --results-dir data/ccp

Reads val_results/ and test_results/ under data/ccp/ and saves a calibration curve to data/ccp/calibration.pdf.

---

Option B — Use pre-trained checkpoints

Step 1. Download data

uv run python scripts/download_data.py --subset ccp

Step 2. Download pre-trained checkpoints

# Download all checkpoints
uv run python scripts/download_models.py

# Or download a specific checkpoint
uv run python scripts/download_models.py --subset ccp

Checkpoints are saved to checkpoints/<subset>/best_model.pth by default.

Step 3. Run inference

uv run python scripts/infer.py ccp --checkpoint checkpoints/ccp/best_model.pth

Step 4. Compute metrics

# Evaluate the current local inference outputs
uv run python scripts/metrics.py ccp

# Or evaluate the exact paper result stacks from Zenodo
uv run python scripts/metrics.py ccp --paper-results

Step 5. (Optional) Calibration

uv run python scripts/calibrate.py ccp --results-dir data/ccp

⚠️ Manual Download of Datasets and Pre-trained Models

If downloading the datasets or pretrained models using the provided script results in an error, you can download them manually from the following link:

Datasets and pretrained models: Zenodo download link

After downloading, extract the files and place them in the appropriate dataset and model directories used by ResMatching.

Model file names

The downloaded best-model checkpoints may include the dataset name as a prefix. For example:

ccp_best_model.pth

Before placing a checkpoint in its corresponding model directory, remove the dataset-name prefix so that the file is named:

best_model.pth

For example:

ccp_best_model.pth  →  best_model.pth

Place each renamed checkpoint inside the directory corresponding to that dataset. Do not place all renamed best_model.pth files in the same directory, as they would overwrite one another.

Similarly, place each manually downloaded dataset in the appropriate dataset directory expected by the project.

---

Citation

If you find this work useful in your research, please consider citing:

@article{resmatching2025,
  title={ResMatching: Noise-Resilient Computational Super-Resolution via Guided Conditional Flow Matching},
  author={Anirban Ray and Vera Galinova and Florian Jug},
  journal={2026 IEEE 23rd International Symposium on Biomedical Imaging (ISBI)},
  year={2026}
}

License

MIT