Logo
  • About
    • Method
    • Convolution modes
      • In a nutshell
      • 2.5D mode
      • 3D mode
    • Training
    • Inference
    • References
  • Install
    • Installing from source
      • One-command environment setup (APS machines, linux-64)
      • Manual environment setup
    • Test the installation
    • Configuration
    • Update
    • Dependencies
  • Usage
    • Data preparation
    • Choosing a convolution mode
      • denoise prepare
      • Sub-reconstructions with tomocupy
    • Training
      • denoise train
        • Running multiple training jobs on the same node
      • What the model learns
      • When to reuse vs. retrain
      • Fine-tuning a pre-trained model
      • Resuming interrupted training
      • Reducing training slices with z_stride
      • Early stopping
      • 3D mode configuration
    • Inference
      • denoise slice (2.5D only)
      • denoise volume
        • Performance example
    • Model Registry
      • Instrument metadata in the config
      • Registering a trained model
      • Searching the registry
      • Using a registered model for inference
    • Command Reference
  • API reference
    • denoise.train
      • count_parameters()
      • run()
    • denoise.slice
      • run()
    • denoise.volume
      • run()
    • denoise.model
      • unet_bottleneck_gn
        • unet_bottleneck_gn.forward()
      • unet_box_gn
        • unet_box_gn.forward()
      • unet_down
        • unet_down.forward()
      • unet_ns_gn
        • unet_ns_gn.forward()
      • unet_up
        • unet_up.forward()
    • denoise.model3d
      • UNet
        • UNet.forward()
        • UNet.forward_gradcp()
        • UNet.weight_init()
      • unet3d()
    • denoise.data
      • PatchIndex
        • PatchIndex.d_idx
        • PatchIndex.left
        • PatchIndex.top
      • TilingMeta
        • TilingMeta.D
        • TilingMeta.H_in
        • TilingMeta.H_pad
        • TilingMeta.P_per_slice
        • TilingMeta.W_in
        • TilingMeta.W_pad
        • TilingMeta.edge_mode
        • TilingMeta.n_cols
        • TilingMeta.n_rows
        • TilingMeta.neighbors
        • TilingMeta.pad_bottom
        • TilingMeta.pad_mode
        • TilingMeta.pad_right
        • TilingMeta.ph
        • TilingMeta.pw
        • TilingMeta.stride_h
        • TilingMeta.stride_w
      • TomoDatasetInfer
        • TomoDatasetInfer.stitch_predictions()
      • TomoDatasetTrain
      • save_normalization_value()
    • denoise.data3d
      • TomoDataset3DInfer
        • TomoDataset3DInfer.stitch_predictions()
      • TomoDataset3DTrain
      • geom_transform_3d()
      • save_normalization_value_3d()
    • denoise.data_utils
      • InferenceBatchSizeOptimizer
        • InferenceBatchSizeOptimizer.estimate_peak_memory()
        • InferenceBatchSizeOptimizer.find_optimal_batch_size()
        • InferenceBatchSizeOptimizer.get_available_memory()
        • InferenceBatchSizeOptimizer.profile()
      • extract_sliding_window_patches_25d()
      • stitch_sliding_window_patches()
      • stitch_sliding_window_patches_core()
    • denoise.loss
      • LCL
        • LCL.forward()
      • laplacian_batch()
      • laplacian_entropy_map()
    • denoise.eval
      • laplacian_batch()
      • laplacian_entropy_map()
      • laplacian_score_batch()
    • denoise.registry
      • list_registry()
      • register()
      • search()
    • denoise.tiffs
      • glob()
      • load_sino()
      • load_stack()
      • natural_sorted()
      • save_stack()
    • denoise.utils
      • save2img()
      • save2img_rgb()
      • scale2uint8()
      • str2bool()
  • Credits
    • Citations
      • Reference
denoise
  • Search

© Copyright 2025, Argonne National Laboratory.

Built with Sphinx using a theme provided by Read the Docs.