Open YJonmo opened 3 weeks ago
Hi @florischabert,
Thank you for reaching out.
In the case of the mentioned random_resized_crop
operator, the default interpolation type is linear
, please use interp_type
parameter and set it to INTERP_NN
.
Amazing, thank you.
One question that is irrelevant to this topic is how could I do this using DALI: The pixels containing values 25 should become 2, and 200 should become 1. The numpy equivalent for a greyscale image is:
images[images==25] = 2
images[images==200]= 1
Do I need to create a custom function and call it in the pipeline as suggested by GPT?
Hi @YJonmo,
I think you can try out the lookup_table
operator:
images= fn.lookup_table(
images,
keys=[0, 25, 100, 220],
values=[0, 2, 100, 1],
)
Describe the question.
Thanks for this work.
I have a pipeline for training image segmentation modes. I am using albumentations library for data augmentations. Now it is time for me to try DALI to get a speed boost.
I need to load the images and corresponding masks and train a model using augmented images and masks. However, I noticed when I Ioad the masks using the following lines the aliasing happens and new pixel values appear in the mask. In other words, the pixel value in the mask frame should be only a certain number such as 0, 25, 100, 220. But after operations such as
fn.resize
orfn.random_resized_crop
values in between them appear.Before resize:![image](https://github.com/NVIDIA/DALI/assets/12040950/efc4e159-34ee-4862-bef1-864ff5cc9f4f)
After resize:![image](https://github.com/NVIDIA/DALI/assets/12040950/1c6f8108-738f-473f-9957-78f8aa337fb3)
So how can I avoid this?
Check for duplicates