Utilities API Reference

Utility functions and helper tools.

Image Processing Utilities

batch_psnr

deeplens.utils.batch_psnr(pred, target, max_val=1.0, eps=1e-8)

Calculate PSNR between image batches.

Parameters:

  • pred – Predicted image batch [B, C, H, W]

  • target – Target image batch [B, C, H, W]

  • max_val – Maximum pixel value

  • eps – Small constant for numerical stability

Returns:

PSNR value in dB

batch_ssim

deeplens.utils.batch_ssim(img, img_clean)

Calculate SSIM between image batches.

Parameters:

  • img – Input image batch [B, C, H, W]

  • img_clean – Reference image batch [B, C, H, W]

Returns:

SSIM value [0, 1]

batch_LPIPS

deeplens.utils.batch_LPIPS(img, img_clean)

Compute LPIPS loss for image batch.

Parameters:

  • img – Input image batch

  • img_clean – Reference image batch

Returns:

LPIPS distance

img2batch

deeplens.utils.img2batch(img)

Convert image to batch format.

Parameters:

img – Image tensor (H, W, C) or (C, H, W), or numpy array

Returns:

Batched image [1, C, H, W]

Image Normalization

deeplens.utils.normalize_ImageNet(batch)

Normalize dataset by ImageNet statistics.

Parameters:

batch – Input image batch

Returns:

Normalized batch

deeplens.utils.denormalize_ImageNet(batch)

Convert normalized images back to original range.

Parameters:

batch – Normalized batch

Returns:

Denormalized batch

Interpolation

interp1d

deeplens.utils.interp1d(query, key, value, mode='linear')

Interpolate 1D query points to the key points.

Parameters:

  • query – Query points [N, 1]

  • key – Key points [M, 1]

  • value – Value at key points [M, …]

  • mode – Interpolation mode

Returns:

Interpolated value [N, …]

grid_sample_xy

deeplens.utils.grid_sample_xy(input, grid_xy, mode='bilinear', padding_mode='zeros', align_corners=False)

Grid sample using xy-coordinate grid [-1, 1].

Parameters:

  • input – Input tensor [B, C, H, W]

  • grid_xy – Grid xy coordinates [B, H, W, 2]

Returns:

Sampled tensor [B, C, H, W]

Video Utilities

create_video_from_images

deeplens.utils.create_video_from_images(image_folder, output_video_path, fps=30)

Create a video from a folder of images.

Parameters:

  • image_folder – Path to folder containing images

  • output_video_path – Output video path

  • fps – Frames per second

Logging and Setup

set_logger

deeplens.utils.set_logger(dir='./')

Setup logger.

Parameters:

dir – Log directory

gpu_init

deeplens.utils.gpu_init(gpu=0)

Initialize device and data type.

Parameters:

gpu – GPU index

Returns:

torch.device

set_seed

deeplens.utils.set_seed(seed=0)

Set random seed for reproducibility.

Parameters:

seed – Random seed

Constants

DEFAULT_WAVE

Default wavelength (0.58756180 um)

WAVE_RGB

RGB wavelengths [0.65627250, 0.58756180, 0.48613270] um

SPP_PSF

Default samples per pixel for PSF (16384)

SPP_COHERENT

Samples per pixel for coherent calculation (~16.7M = 2^24)

SPP_CALC

Samples for computation (1024)

SPP_RENDER

Samples per pixel for rendering (32)

DEPTH

Default object depth (-20000.0 mm)

PSF_KS

Default kernel size for PSF calculation (64)

EPSILON

Small constant to avoid division by zero (1e-9)