Boilerplate free Image toolkit for PyTorch notebooks. Make one-liners like this.
I.search_images("Cats & Dogs").show()
I.search_history().crop(224).pt()
I.download(some_image_urls).crop(224).show()
I.download(some_image_urls).to_dir("/path/to/images")
pytorch_images = I.from_dir("path/to/images").resize(256).pt()
images = I.wrap(pytorch_images ).crop(224)
images.show()
pil_images = images.pick(0, 3, 4).normalize().pil()
I.merge(images, pil_images, pytorch_images, pil_images[0]).resize(128).to_dir("thumbs")
- Play with the Tutorial: Colab, GitHub
- Watch the Tutorial
- Bring in images from all sources into a single place like PyTorch images, PIL images, Disk, or Internet
- Apply various transformations like crop, resize, normalization, etc
- Then output them as you like including displaying, PyTorch tensors, inline video, and to disk.
- Imagine all those things without some boilerplate free code and one-liners
!rm -rf ./imitools && git clone https://github.com/GDi4K/imitools.git
!pip install -q duckduckgo_search
import imitools as I
Now you can use all the utility functions via the I
global namespace.
For more usage, check the tutorial: Colab, GitHub
You can also look at the source code for more info.