Observations
Must not forget that Detectron2, Pytorch-lightning, Hugging-face transformers, and Pytorch has option to resume training.
With this resume training option, we can continuously train on colab.
image.shape
works in opencv
- outputs (height, width, channels)
image.size
works in PIL(pillow)
- outputs (width,height).
- img.mode gives no.of.channels like RGB
.
then what is torch follows..?
if !cd
doesn't work in colab, then try %cd
. check here
Shell command in python code: