Lesson3-head-pose.ipynb: Creating a dataset+

Part 1 (2019) Advanced (Part 1 v3)
1

Hi, when running the below code to create a dataset and get such errors, how to fix this issue/error? thanks.

Bintao Li

The code:
data = (PointsItemList.from_folder(path)

    .split_by_valid_func(lambda o: o.parent.name=='13')

    .label_from_func(get_ctr)

    .transform(get_transforms(), tfm_y=True, size=(120,160))

    .databunch().normalize(imagenet_stats)

   )

The errors:
Exception: It’s not possible to apply those transforms to your dataset:

Expected object of scalar type Double but got scalar type Float for argument #3 ‘mat2’ in call to _th_addmm_out

2

Hi, bintao li
Try to change the

transform(get_transforms(), tfm_y=True, size=(120,160))

to be:

transform(get_transforms(), tfm_y=True, size=(120,120))

If the thing in above does not work, try to change it to be size=(60, 60)

3

Hi JonathanSum, thanks for response. However, it does not work for both size(120,120) and (60,60)(for the same issue). There is something wrong with the source code?

Best,

Bintao li

4

Here is mine that works on colab:
https://colab.research.google.com/drive/1TNv8pSnNW_iH3e9kQk9XQj0_n9wzF8C3?usp=sharing
lbt, if you really want people to help you, please share your full code( reproduction code that has your error). I will try to run them on colab and try to find out the error in my free time.

Here is what I used:

data = (PointsItemList.from_folder(path)
    .split_by_valid_func(lambda o: o.parent.name=='13')
    .label_from_func(get_ctr)
    .transform(get_transforms(), tfm_y=True, size=(120,160))
    .databunch().normalize(imagenet_stats)
   )
5

Hi JonathanSum, thanks!
My strategy is to run all notebooks of Part 2 (to have a big picture/feeling about it) and then come back to figure out some errors/issues.

Sorry for my late response.
Please check the sharing link for you reference:
https://colab.research.google.com/drive/1hRWIcDBlpJ3PKrg-G-UZp78YTHqCLlix?usp=sharing