vodet.utils module¶
- vodet.utils.discrete_cmap(N, base_cmap=None)¶
Create an N-bin discrete colormap from the specified input map
- vodet.utils.exif_date(in_dir)¶
Read Exif data of images in a directory and returns a dataframe of file names of images and their shooting date.
- Parameters
in_dir (str) – Path for the directory of images.
- Returns
df – A dataframe of filename and shooting date.
- Return type
pd.DataFrame
- vodet.utils.make_patches_labelled(data_dir, out_dir, label_data, step_ratio)¶
Split source images with labels.
- Parameters
data_dir (str) – The path for data directory.
out_dir (str) – The path for output (patched images) directory.
step_ratio (float) – Sliding window step size relative to the size of patches.
- vodet.utils.make_patches_unlabelled(data_dir, out_dir, label_data, step_ratio)¶
Split source images into patches. The patch sizes are determined by label_data.
- Parameters
data_dir (str) – The path for data directory.
out_dir (str) – The path for output (patched images) directory.
label_data (pd.DataFrame) – A dataframe of label data. This will be used to determine the sizes of patches.
step_ratio (float) – Sliding window step size relative to the size of patches.
- vodet.utils.plot_latent(x, y, f, q, y_dim)¶
- vodet.utils.plot_reconstruction(x, y, p, q)¶
- vodet.utils.read_labelme(in_dir)¶
Read labelme’s json files and convert it into pd.DataFrame
- Parameters
in_dir (str) – The directory where json files are stored.
- Returns
dataframe
- Return type
pd.DataFrame
- vodet.utils.set_patches(data_dirs, label_type, step_ratio=1.0)¶
Making labelled and unlabelled patches from annotated images. The annotation shape must be rectangulars. Only compatible with VoTT (csv export) and labellme.
- Parameters
data_dirs (dict) – A dictionary with 3 components, “train”, “validation” and “unlabelled” . The “train” and “validation” directory should have - /source: a sub-directory with source images - /labels: a sub-directory with label data. label data should be VoTT csv-export file (.csv) or labelme output file (.json)
label_type (str) – The type of label files, “VoTT” and “labelme” are available.
step_ratio (float default 1.0) – Sliding window step size relative to the size of patches.