instances mask and category

[chengzheng-nreal]

Thank you very much for the open source dataset, there is a problem, I found that there is only instance information in the instance_id_maps folder, the instance mask is 0, how do I get the category and mask corresponding to each instance?

[nikitakaraevv]

Hi @chengzheng-nreal, thanks for reaching out!
Your observation is indeed correct. At first glance, instance masks may appear empty because they are encoded as integers starting from 0. However, this is not the case.

Instance masks can be visualized by multiplying their values by a certain factor. In the case of three instances (0,1,2), you can multiply by 128.

You can try the following code snippet to visualize instance masks:

from PIL import Image
import matplotlib.pyplot as plt
import numpy as np

img_path = './dynamic_replica/test/cb2501-3_obj_source_right/images/cb2501-3_obj_source_right-0899.png'
instance_map_path = img_path.replace('images','instance_id_maps').replace('right-','right_')

img = Image.open(img_path)
instance_map = Image.open(instance_map_path)

plt.imshow(img)
plt.imshow(np.array(instance_map)*128)

Does it solve your problem?

[chengzheng-nreal]

Hi @chengzheng-nreal, thanks for reaching out! Your observation is indeed correct. At first glance, instance masks may appear empty because they are encoded as integers starting from 0. However, this is not the case.

Instance masks can be visualized by multiplying their values by a certain factor. In the case of three instances (0,1,2), you can multiply by 128.

You can try the following code snippet to visualize instance masks:

from PIL import Image
import matplotlib.pyplot as plt
import numpy as np

img_path = './dynamic_replica/test/cb2501-3_obj_source_right/images/cb2501-3_obj_source_right-0899.png'
instance_map_path = img_path.replace('images','instance_id_maps').replace('right-','right_')

img = Image.open(img_path)
instance_map = Image.open(instance_map_path)

plt.imshow(img)
plt.imshow(np.array(instance_map)*128)

Does it solve your problem?

Thank you very much for your reply, which has solved most of my problems, I found that I made a mistake earlier cv2.imread read uint8, should read uint16. Now another question is how do I know which instances are people, I want to automatically filter out the instances of people