meyert11 / travnet

TravNet is a project towards the full automation of spike sorting, the identification of neurons from noise. This is achieved by training neural networks on sorts completed by human experts.

🤗 Models on Hugging Face 

Introducing TravNet

TravNet is a suite of functions to fully automate the seperation of neurons from noise. It uses preprocessing to take a binary data file and convert it to a waveblob (waveforms and their corresponding principal components). The waveblobs are then sorted using a pretrained convolutional neural network, trained on human sorted batches. The output is a file containing the waveforms corresponding cluster ID (neuron template), and timestamp.

Quick Start

You can follow the steps below to quickly get up and running with TravNet models.

1. Create a conda env

2. In the top-level directory run:

   pip install -e .

3. To download model weights and sample data:

   python3 download_samples.py

4. Once the model weights and data are available you can run the model using the command below:

   python3 example_sorter.py