WolfDetector is a visual object detection library focusing on ease of use and accessibility. It is implemented in the Wolfram Language. The only thing you need to install to get started is the Wolfram Engine (currently a 1GB download, cross-platform).
WolfDetector is based on the YOLOv2 neural network architecture, and it uses a pre-trained model to speed up training time and reduce the number of training examples you need for your dataset. Because of this, training can be done quickly with only a laptop CPU.
Additionally, WolfDetector can take advantage of Nvidia GPU technology to make training even faster.
Object detection the problem of locating objects in an image. WolfDetector is a program which learns to do object detection by learning from data. To train WolfDetector on your data, you first need to build your own dataset - or alternatively find one that someone else has made.
After you have a dataset, it is simple to use the environment of your choice to complete training and run your WolfDetector.
-
Get your dataset
-
Extract key information from that dataset
-
Create a training model for your dataset
-
Convert your dataset to a training dataset
-
Train your model
-
Convert your trained model to a prediction model
-
Use your prediction model 🚀
TODO
TODO
- Wolfram Language Library ( 2021/06/24 )
- Wolfram Language Paclet
- Wolfram Function Repository
- MXNet
- ONNYX
- TensorFlow
- PyTorch
- Pascal VOC style dataset ( 2021/06/29 )
- MSCOCO style dataset
- Darknet style dataset
- Python Bindings
- Scripts (Wolframscript)
- Java Bindings
- C# Bindings
- C Bindings