Ok I have thanks to @willismar fastai up and running on a Docker image, I’ll probably install also on my desktop. Really basic knowledge of Python, basic experience of image classification using commercial services such as Watson and Wize.
I have a very specific goal, to prototype a image classifier for a specific sector and just learn the how to and elements that can influence the final result and enable me planning a projects.
Want to ask if I need in your opinion to study all the fastai material or there is some easy start to try and experiment with.
Also a sub-doubt is if using Docker will substantially reduce the performances, or there are configuration to achieve near native installation speed.
You can achieve this by watching the first 2-3 videos of Part 1v2, using fastai v0.7, and working with the associated notebooks. The concepts taught here are transferable and probably current enough for your needs.
In a month or two, Part 1v3, using fastai 1.0, will be generally available and early lectures also concern image classification.
Thanks @digitalspecialists
Currently installed fastai 1 on Docker, searching I didn’t find any available docker image with fastai 0.7, wondering if is possible to downgrade to 0.7 from 1.0, or I need to recreate the whole environment using 0.7.
Also 1.0 is so different from 0.7 to have problems using current lessons, or can I discover the differences without many troubles.
a sub-doubt is if using Docker will substantially reduce the performances, or there are configuration to achieve near native installation speed.
Let me know your experience using it.
Also 1.0 is so different from 0.7 to have problems using current lessons, or can I discover the differences without many troubles.
Here is the documentation for fastai v1.0, there is a session with tutorials to follow. https://docs.fast.ai/
I didn’t find any available docker image with fastai 0.7, wondering if is possible to downgrade to 0.7 from 1.0, or I need to recreate the whole environment using 0.7.