Is there are a way to read images into a prediction engine without saving to file first?

Question

All the of the Microsoft tutorials on ML.NET involve reading images from a file and referencing them only using their file location. I'm trying to implement a prediction engine into piece of code which is taking photos as it is being run. Currently I'm saving all my images to file and then reading them to make a prediction but I'm wondering if there's a better way?

What is the API documentation for ML.NET saying? Look at the type that offers the file load method you are currently using. Chances are it also offers other load methods for loading data from a stream or similar... — , Jun 28 '19 at 18:31
[Looks like](https://github.com/dotnet/machinelearning/issues/3839) it's in their backlog :) — Jon, Jun 28 '19 at 18:41

Neil N · Answer 1 · 2019-06-28T18:47:48.173

0

you should be able to use a MemoryStream in place of a FileStream anywhere a file is needed

edited Jun 28 '19 at 18:47

answered Jun 28 '19 at 18:44

Neil N

24,862
16
85
145

it's a comment ? – A Farmanbar Jun 28 '19 at 18:46

score 0 · Accepted Answer · answered Jul 08 '19 at 18:07

This was enabled in MAdded support for in-memory ‘image type’ in IDataview was enabled in ML.NET 1.1. Please read this blog post for more details https://devblogs.microsoft.com/dotnet/announcing-ml-net-1-1-and-model-builder-updates-machine-learning-for-net/

Is there are a way to read images into a prediction engine without saving to file first?

2 Answers2