Bug fixes in read_image_to_tensor

[stereomatchingkiss]

1. Select cpu if gpu not available
2. Change BGR to RGB before inference, this will fix some bugs in Add a dataloader as ultralytics in detection pipeline #92
3. Append sys path, so the notebook can open on windows

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[CLAassistant]

All committers have signed the CLA.

[codecov]

Codecov Report

Merging #97 (bfd0362) into master (dbe2d4d) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master      #97   +/-   ##
=======================================
  Coverage   83.39%   83.39%           
=======================================
  Files           8        8           
  Lines         777      777           
=======================================
  Hits          648      648           
  Misses        129      129           

Flag	Coverage Δ
unittests	83.39% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update dbe2d4d...bfd0362. Read the comment docs.

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

zhiqwang commented on 2021-04-21T04:49:49Z
----------------------------------------------------------------

Actually we assume that the notebooks are decoupled with the path dependence, in other words, we assume this notebook could be working in any path, so I don't recommend adding sys path dependency here. (I am plan to add these notebooks to colab or kaggle, with some python library added, like pip install -U yolort) And this is the reason why I add get_image_from_url here, the image are downloaded from the internet.

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

zhiqwang commented on 2021-04-21T04:49:50Z
----------------------------------------------------------------

Hi @stereomatchingkiss , I think that you could move the BGR2RGB operation to read_image_to_tensor directly.

https://github.com/zhiqwang/yolov5-rt-stack/blob/dbe2d4d1dd320be34d4c904e6cf6582ea91e868f/yolort/utils/image_utils.py#L167-L183

[stereomatchingkiss]

Thanks for your helps, will fix

View / edit / reply to this conversation on ReviewNB

zhiqwang commented on 2021-04-21T04:49:50Z

Hi @stereomatchingkiss , I think that you could move the BGR2RGB operation to read_image_to_tensor directly.

https://github.com/zhiqwang/yolov5-rt-stack/blob/dbe2d4d1dd320be34d4c904e6cf6582ea91e868f/yolort/utils/image_utils.py#L167-L183

Thanks for your help, fixed

[zhiqwang]

LGTM!

[zhiqwang]

Hi @stereomatchingkiss , Thanks for your feedbacks and contributions here, welcome to join yolort community.

Besides, it would be more better if you could sign the CLA at .