Getting better results¶
Skeleton design answers some questions about designing the skeleton for your animals.
Choosing a set of models provides information about the types of models you should.
Add more training data to a project when you have predictions that aren’t in the same project as your original training data and you want to correct some of the predictions and use these corrections to train a better model.
Tracking and proofreading provides tips and tools you can use to speed up proofreading when you’re happy enough with the frame-by-frame predictions but you need to correct the identities tracked across frames.
Export a training package when you’ve created a project with training data on one computer and you want to move your training data for training models on a different computer.
Run training and/or inference on Colab when you have a project with labeled training data and you’d like to run training or inference in a Colab notebook.
Remote training when you have a project with training data and you want to train on a different machine using a command-line interface.
Creating a custom training profile for creating custom training profiles (i.e., non-default model hyperparameters) from the GUI.
Remote inference when you trained models and you want to run inference on a different machine using a command-line interface.
Downloading pretrained weights for remote training explains how to download pretrained weights if you’re training a network with pretrained weights on a machine without internet access.
- Choosing a set of models
- Run training and/or inference on Colab
- Creating a custom training profile
- Add more training data to a project
- Tracking and proofreading
- Feature Reference
- Remote training
- Downloading pretrained weights for remote training
- Remote inference
- Skeleton design
- Training with GUI
- Export a training package