<aside> <img src="/icons/warning_gray.svg" alt="/icons/warning_gray.svg" width="40px" /> The following guide is created on macOS so you might need to adapt a few of them on your platform. Feel free to ask me questions via email.

</aside>

<aside> <img src="/icons/light-bulb_gray.svg" alt="/icons/light-bulb_gray.svg" width="40px" /> Follow previous guide (Export a Labelme JSON file to PNG and etc. ) to export files or download them from below:

<aside> <img src="/icons/document_gray.svg" alt="/icons/document_gray.svg" width="40px" /> The exported files that are zipped (to open please double-click to extract):

dogs_edited_json.zip

</aside>

</aside>

1. Make sure you have dogs_edited_json directory.

   $ ls dogs_edited_json
   img.png  label.png  label_names.txt  label_viz.png
   

2. Install required software following Install labelme Python package

3. Create a Python file load_exported_files.py:

   import numpy as np
   import PIL.Image
   
   import matplotlib.pyplot as plt
   
   # load exported files
   image = np.asarray(PIL.Image.open("dogs_edited_json/img.png"))
   label = np.asarray(PIL.Image.open("dogs_edited_json/label.png"))
   with open("dogs_edited_json/label_names.txt", "r") as f:
       label_names = f.read().splitlines()
   
   # extract masks from label
   masks = {}
   for label_id, label_name in enumerate(label_names):
       mask = label == label_id
       masks[(label_id, label_name)] = mask
   
   # print stats
   print("image:", image.shape, image.dtype)
   print("label:", label.shape, label.dtype)
   print("label_names:", label_names)
   
   # visualize
   rows = 2
   columns = max(2, len(label_names))
   #
   plt.subplot(rows, columns, 1)
   plt.title("image")
   plt.imshow(image)
   #
   plt.subplot(rows, columns, 2)
   plt.title("label")
   plt.imshow(label)
   #
   plt.subplot(rows, columns, 3)
   plt.title("label overlaid")
   plt.imshow(image)
   plt.imshow(label, alpha=0.5)
   #
   for (label_id, label_name), mask in masks.items():
       plt.subplot(rows, columns, 4 + label_id)
       plt.title(f"{label_id}:{label_name}")
       plt.imshow(mask, cmap="gray")
   #
   plt.tight_layout()
   plt.show()
   

   

   

   <aside> <img src="/icons/document_gray.svg" alt="/icons/document_gray.svg" width="40px" /> The Python file:

   load_exported_files.py

   </aside>

That’s all! Now you’re ready to start building your own dataset using Labelme.