Skip to content

prismpyp eval2d

Purpose

Evaluate and visualize 2D embeddings or model features from trained SimSiam networks.

This command can:

  • Run 2D embedding evaluation on real-space or Fourier-space inputs
  • Optionally resume from precomputed embeddings
  • Perform clustering and UMAP dimensionality reduction
  • Save visualizations, plots, and .webp image thumbnails

Usage

usage: prismpyp eval2d [-h] [--output-path DIR] [--metadata-path METADATA_PATH] [--embedding-path [EMBEDDING_PATH]] [-a ARCH] [-j N] [--epochs N]
                       [--start-epoch N] [-b N] [--lr LR] [--momentum M] [--wd W] [-p N] [--resume PATH] [--feature-extractor-weights PATH]
                       [--world-size WORLD_SIZE] [--rank RANK] [--dist-url DIST_URL] [--dist-backend DIST_BACKEND] [--seed SEED] [--gpu GPU]
                       [--multiprocessing-distributed] [--dim DIM] [--pred-dim PRED_DIM] [--fix-pred-lr] [--use-fft] [--downsample DOWNSAMPLE]
                       [--pixel-size PIXEL_SIZE] [--size SIZE] [--evaluate] [--n-clusters N_CLUSTERS] [--num-neighbors NUM_NEIGHBORS]
                       [--min-dist-umap MIN_DIST_UMAP] [--n-components N_COMPONENTS] [--nextpyp-preproc NEXTPYP_PREPROC] [--zip-images]

Named Arguments

Commonly Changed

Argument Description Default
--output-path DIR Path to output directory
--metadata-path METADATA_PATH Path to metadata file
--embedding-path EMBEDDING_PATH Optional path to precomputed embeddings
--feature-extractor-weights PATH Path to pre-trained feature extractor weights none
--dim DIM Feature dimension 512
--pred-dim PRED_DIM Hidden dimension of the predictor 256
--fix-pred-lr Fix learning rate for the predictor True
--use-fft Use FFT of the image as input False
--evaluate Evaluate model on validation set True
--n-clusters N_CLUSTERS Number of clusters for KMeans
--num-neighbors NUM_NEIGHBORS Number of neighbors for UMAP
--min-dist-umap MIN_DIST_UMAP Minimum distance for UMAP 0
--n-components N_COMPONENTS Number of UMAP components

Available Architectures:
resnet18, resnet34, resnet50