search

ludwig-ai / ludwig

Low-code framework for building custom LLMs, neural networks, and other AI models
http://ludwig.ai
Apache License 2.0
11.18k stars 1.19k forks source link

Image Decoder #679

Closed jplehmann closed 3 weeks ago

jplehmann commented 4 years ago

I know this is a feature being worked on, but I didn't see an issue for it.

E.g., I'd like to build a stacked CNN that removes noise from photos.

Any sense of how long until this is available? Thanks!

w4nderlust commented 4 years ago

@jplehmann thank you for opening this, it is indeed something we have in the backlog, but we are currently focusing on other aspects (see https://github.com/uber/ludwig/projects/1 , the image output features are in the triage section). It's difficult to give a timeline for this. It will likely not be in v0.3 which will likely come out in the next two months, but it is likely that image outputs, together with other outputs features (bounding box for object detection and timeseries) will be prioritize when the works for v0.4 begin. As for any other feature, we welcome contribution, I can provide full guidance on how to go and implement the feature if someone is willing to help.

jplehmann commented 4 years ago

@w4nderlust I'm probably not the person to build this out right now, but thank you for the quick response (again) and info. Really excited about this project.

jeffin07 commented 4 years ago

@w4nderlust i would like to work on this issue

w4nderlust commented 4 years ago

@jeffin07 it would be great, but I would push it to after when we are done with TF2 porting, as this will require adding some modeling code.

jeffin07 commented 4 years ago

@w4nderlust okay hope to work on this after TF2 porting

qiagu commented 2 years ago

Any update on this?