stratus2019/score_image_with_cnn 0.0.1

alg install stratus2019/score_image_with_cnn
score image with cnn
Algorithm Description

Scores an image given a convolutional neural network using a sliding windowed approach. The user can specify the window size (in pixels), the target chip size (image size used by the trained neural network), and percent overlap to use for the sliding window, etc. The algorithm generates a colormapped image, which is saved in the output directory specified by the user. It also returns image scores and the upper left hand coordinates of the corresponding image chip windows for further downstream processing.




Name Description Data Type Required?
base_data_dir base directory for data string
base_save_dir None string
color_palette None string
image_data None string
overlap_percent_0_100 None string
relative_model_dir relative directory to cnn model, starting from the base data dir string
relative_save_dir None string
thing_to_find None string
window_size_npix None string
Name Description Data Type
scores string
upper_lefts string

  • 0.0.1 (current)