Denoising¶

The denoising module reduces noise in cryo-electron tomograms while preserving structural detail. It is typically the first step in a CryoSiam pipeline, preceding semantic segmentation, instance segmentation, or particle identification.

For best performance, it is recommended to use WBP-reconstructed tomograms, as the denoising model was trained on simulated WBP data.

Left: Raw WBP noisy tomogram
Middle: Gaussian filtered tomogram
Right: CryoSiam denoised tomogram

Trained Model¶

A pre-trained denoising model is provided and can be downloaded from:

CryoSiam denoising model (v1.0)

Running denoising¶

Denoising is performed via the command-line interface using a YAML configuration file:

cryosiam denoise_predict --config_file=configs/config_denoising.yaml

What this command does

Loads the trained denoising model and the input tomogram(s)
Applies sliding-window 3D inference (GPU)
Writes denoised tomograms to the specified output directory
Optionally saves intermediate or raw prediction files

Visualization Denoised outputs can be visualized using CryoSiam-Vis. See the visualization instructions here.

Example Configuration File¶

Download example config

data_folder: '/scratch/stojanov/dataset1/tomograms'
prediction_folder: '/scratch/stojanov/dataset1/predictions/denoised'

trained_model: '/scratch/stojanov/trained_models/cryosiam_denoising.ckpt'
file_extension: '.mrc'

test_files: null

save_raw_predictions: False
scale_prediction: True

parameters:
  data:
    patch_size: [ 128, 128, 128 ]
    min: 0
    max: 1
    mean: 0
    std: 1
  network:
    in_channels: 1
    spatial_dims: 3

hyper_parameters:
  batch_size: 3

Config Reference¶

Top‑level keys¶

Key	Type	Must change the default value	Description
`data_folder`	`str`	✅	Directory containing input tomograms to denoise
`prediction_folder`	`str`	✅	Output directory for denoised tomograms
`trained_model`	`str`	✅	Path to the trained denoising model checkpoint (`.ckpt`)
`file_extension`	`str`	❌	Input file extension (`.mrc` or `.rec`, default: `.mrc`)
`test_files`	`list[str]` or `null`	❌	List of specific filenames to process. Use `null` to process all files in `data_folder`
`save_raw_predictions`	`bool`	❌	Save raw prediction files (.h5) in addition to final outputs
`scale_prediction`	`bool`	❌	Rescale output intensities to [0, 1]

`parameters`¶

Key	Type	Must change the default value	Description
`data.patch_size`	`list[int]`	❌	Sliding-window patch size for 3D inference
`data.min`	`float`	❌	Intensity minimum value for data scaling
`data.max`	`float`	❌	Intensity maximum value for data scaling
`data.mean`	`float`	❌	Mean used for normalization
`data.std`	`float`	❌	Std used for normalization
`network.in_channels`	`int`	❌	Number of input channels (usually `1`)
`network.spatial_dims`	`int`	❌	Dimensionality of the model (`3` for tomograms)

Tips
Keep patch_size consistent with the value used during training (e.g. 128³ for the provided model). Reduce it only if you encounter GPU memory issues.

`hyper_parameters`¶

Key	Type	Must change the default value	Description
`batch_size`	`int`	❌	Number of 3D patches processed per forward pass.

Outputs¶

The denoising step produces:

Denoised tomograms saved to prediction_folder (file extension as implemented).
Optional raw prediction files if save_raw_predictions: true.

Output filenames follow the input basenames with appropriate suffixes.

Troubleshooting¶

Symptom	Suggested Fix
`CUDA out of memory`	Reduce `batch_size` and/or `patch_size`; ensure no other heavy GPU processes are running.
No files processed	Ensure `data_folder` exists and `test_files` is `null` or matches filenames exactly.
Output intensities look off	Toggle `scale_prediction`.
Model/shape mismatch	Verify `in_channels` and `spatial_dims` and the correct `.ckpt` file are used.

Next Steps¶

After denoising, continue with:

For an overview of complete pipelines, see the Usage overview.