Skip to content

04-analysis.md

Latest commit

 

History

History
45 lines (29 loc) · 1.99 KB

04-analysis.md

File metadata and controls

45 lines (29 loc) · 1.99 KB
Raw

Interactive plots

BioEncoder features interactive PCA and T-SNE visualizations so that you can check the embeddings you get after the training. To generate the interactive plot, use:

bioencoder.interactive_plots(config_path=r"bioencoder_configs/plot_stage1.yml")

Interactive plots of the first stage allow the exploration of the embedding space and distances between individuals.

Model explorer

We also provide a model visualization playground, where individuals can get further insight into their data. To launch the app and explore the final classification model, simply use:

bioencoder.model_explorer(config_path=r"bioencoder_configs/explore_stage1.yml")
bioencoder.model_explorer(config_path=r"bioencoder_configs/explore_stage2.yml")

Model visualization techniques vary between first and second stage, so make sure to select the appropriate ones.

Inference

To extract embeddings (stage 1) or to classify (stage 2) individual images using a trained model, supply a config file (to specify the stage and the image size at training) and a path to an image (or a numpy array) to the inference script:

bioencoder.inference(config_path=r"bioencoder_configs/inference.yml", image="path/to/image.jpg")

Note that in CLI mode you can either supply a path to an single image, or to a directory of images. There a few more options - check them with help:

bioencoder_inference --config-path "bioencoder_configs/inference.yml" --path "path/to/image"

optional arguments:
  -h, --help            show this help message and exit
  --config-path CONFIG_PATH		Path to the YAML configuration file to create interactive plots.
  --path PATH          			Path to image or folder to embedd / classify.
  --save-path SAVE_PATH 		Path to CSV file with results.
  --overwrite           		Overwrite CSV file with results.