Output

The pipeline's final output is a map of the plants it detected in your drone images, overlaid on top of the orthomosaic and colored according to the species that the pipeline assigned to them. This file will have the path <output_dir>/<dataset_ID>/map-exp.tiff (or <output_dir>/<dataset_ID>/map-exp.tiff if running the traditional strategy)

However, there are several intermediary directories and files that are also output by the pipeline. Note that the files that you see when you execute the pipeline may be only a subset of this list, depending on how you execute it.

Traditional strategy

• stitch - the output of the stitching step: a high-quality orthomosaic and a Metashape project file
• segments/high.json - high-confidence guesses for the shape of each plant in the orthomosaic
• segments/low.json - low confidence guesses for the shape of each plant in the orthomosaic
• segments.json - the output of the segmentation steps for the orthomosaic: polygons (ie segments) representing the shapes of each plant in the orthomosaic
• features - machine learning features for every segment in the orthomosaic
• truth_data.tsv - truth data for training/testing the classifier
• train - training data and the results of training the classifier: a trained random forest model and the random forest importance of each feature
• test - testing data and the results of testing the classifier
• classify/ortho.tsv - final species classifications for each segment from the orthomosaic
• segments-map.tiff - a visualization of the outputs of the segmentation steps in the pipeline
• map.tiff - a visualization of the outputs of the segmentation and classification steps of the pipeline

Experimental strategy

• stitch-lowQual - the output of the stitching step under the low quality configuration option
• segments/high - high-confidence guesses for the shape of each plant in each image
• segments/low - low confidence guesses for the shape of each plant in each image
• transforms - high and low confidence guesses for the shape of each plant in each image, transformed to orthomosaic coordinates
• segments-exp.json - the output of the segmentation steps for each image: polygons (ie segments) representing the shapes of each plant in the orthomosaic
• rev_transforms - segments of each plant in the coordinates of each image
• features-exp - machine learning features for every segment in every drone image
• truth_data-exp - truth data for testing the classifier
• truth_data-exp.tsv - truth data for training/testing the classifier
• train-exp - training data and the results of training the classifier: a trained random forest model and the random forest importance of each feature
• test-exp - testing data and the results of testing the classifier
• classify - the result of applying the classifier on the segments from each drone image
• results.tsv - final species classifications for each segment
• segments-map-exp.tiff - a visualization of the outputs of the segmentation steps in the pipeline
• map-exp.tiff - a visualization of the outputs of the segmentation and classification steps of the pipeline

Both strategies

• benchmark - contains benchmarking files for every step in the pipeline
• log - the basic snakemake log of completed rules; an abridged summary of the pipeline's progress and any errors
• qlog - a more detailed log of the progress of each rule and the reasons for any errors