LightNet - Python Interface to Darknet [Repo]
Fellows at Explosion.Ai have created a github repository for LightNet, which provides a simple Python interface to DarkNet.
For those who don't know what DarkNet is, this is a neural network library created by Joseph Redmon; it's a neural net popular for its top-notch object detection models from the YOLO family (v1 and v2). YOLO stands for: you only look once, and I guess that if you are familiar with the machine learning field, you know what this alludes to.
As per the team:
"LightNet's main purpose for now is to power Prodigy's upcoming object detection and image segmentation features. However, it may be useful to anyone interested in the DarkNet library." [source]
There are some main advantages for using LightNet, one of them being that it allows you to work with YOLO2 for state of the art object detection. Another advantage is that it's very easy to use with Python, allowing you to pass in byte strings and returning box-bounded NumPy arrays.
I'd say that one of the best features is that it's self-contained and light-weight, in that it's doesn't depend on TensorFlow or other machine learning libraries. It's dependency is DarkNet.
You can install it by simply doing pip install lightnet and python -m lightnet download yolo. However, it currently supports only OSX and Linux and Python 2.7, 3.5 and 3.6. The team says that they are working on a Windows release as well.
Those of you who are not familiar with machine learning and deep learning frameworks, after reading the title of this post, might have thought that this naming of DarkNet could have something to do with shady operations on the 'darkweb' or whatnot. Clearly, Joseph Redmon, lacked a moment of inspiration when looking for a name for his library.
The developers of LightNet are quick into making the disassociation by saying that a feature of LightNet is:
- "Not named DarkNet: Avoids some potentially awkward misunderstandings." [source]
Anyway, if you're on OSX or Linux and you want to play with LightNet, possibly for object-detection, you might want to visit the main repository by following the link below. Here's a teaser for what YOLO2 can do:
To stay in touch with me, follow @cristi
Cristi Vlad Self-Experimenter and Author
great topic ,good luck friend
Wonderful post bro @cristi
Nice post
post a very interesting @cristi, I really liked it, therefore I am back waiting for your post the next. thank you for attend to share the back.
hallo @cristi Your post is very useful for people who are learning python and can add their insight, thanks greetings from me