DLIB Features

[VisionEp1]

Hi, it has been a while. I found some old related questions but i think its more productive to ask it here. Mainly its about "new" AI features and the feature of DLIB in generall.

Do you have any plans ? (or other contributors) to develop one of the following features: And if not so, why not? Or would you be willing to merge any high-quality contribution into dlib regarding those features:

1. Single Shot Detector (AKA YOLO or SSD)
2. 2019 State of the Art Object Detection (for example Cascade RCNN)
3. Weakly Supervised (or semi supervised) learning for object detection? (or classification)
4. Generative Adversarial Networks (GANs) (i know the layers all exist already to my knownledge, but not something like aworking GAN example)

thanks a lot and best wishes. In my opinion dlib is one of the best written ML libaries out there and i would love it seeing getting extend to current topics of AI research.

• Off topic: In case its a matter of ressources, are you willing to accept donation to be able to invest more time into dlib?

best wishes and thx a lot

[davisking]

I don't have any specific plans. But this is nothing new. I've never worked on dlib with any kind of roadmap in mind. Rather, I work on whatever I find interesting to work on at that moment.

I'm always happy to get high quality pull requests though. If you want to implement something from a recent CVPR paper or anything like that and get it to a high quality that would be a great contribution. Any of the things you just mentioned fit that description.

[DlibReza]

Hello Davis Do you support newest version of CUDA when release in future years? i am anxious because you dont release any version after 19.17 version.

[davisking]

The version of dlib on github should work with the newest cuda version.

[davisking]

It's been too long since the last release, so I just made v19.18 :)

[DlibReza]

Hello Davis Thank you for your answer and for new version( v19.18) of Dlib.

[VisionEp1]

Thanks for your awnser. Do you see yourself working on any of those topics (1-4) in the near future? (so in case i do a PR we wouldn't do the same thing)

[davisking]

I’m not going to work on them in the near future. So feel free to tackle any of them.

[VisionEp1]

Thanks one more thing, i know dlib uses cuda.

When i would add a new loss layer, would i need to implement the cuda implementation aswell ?

(didnt work to much with cuda dev. yet)

[davisking]

Maybe. But most of the loss layers in dlib are just cpu code. So unless you are doing something really compute or bandwidth intensive cpu only is fine.

[dlib-issue-bot]

Warning: this issue has been inactive for 35 days and will be automatically closed on 2019-11-08 if there is no further activity.

If you are waiting for a response but haven't received one it's possible your question is somehow inappropriate. E.g. it is off topic, you didn't follow the issue submission instructions, or your question is easily answerable by reading the FAQ, dlib's official compilation instructions, dlib's API documentation, or a Google search.

[dlib-issue-bot]

Warning: this issue has been inactive for 43 days and will be automatically closed on 2019-11-08 if there is no further activity.

If you are waiting for a response but haven't received one it's possible your question is somehow inappropriate. E.g. it is off topic, you didn't follow the issue submission instructions, or your question is easily answerable by reading the FAQ, dlib's official compilation instructions, dlib's API documentation, or a Google search.

[dlib-issue-bot]

Notice: this issue has been closed because it has been inactive for 45 days. You may reopen this issue if it has been closed in error.