Skip to content

Latest commit

 

History

History
27 lines (22 loc) · 1.44 KB

README.md

File metadata and controls

27 lines (22 loc) · 1.44 KB

Automated Neuron Detection using YOLOv3

By Mighten Yip and Mercedes Gonzalez (Precision Biosystems Laboratory at Georgia Tech, 2020). Referenced manuscript published in Scientific Reports.

This repository allows the camera (those compatible with Micro-manager) to interact with the neuron detection YOLOv3 algorithm in acute mouse brain slice. Parts of the code were cloned from /qqwweee/keras-yolo3. Patch clamp experiments to verify the health of the neurons used Scientifica manipulator/stage and a digitally controlled pressure control box.

To get started: clone/fork this repository into your preferred folder location.

live_neuron_detection.py is the origin file to run the neuron detection algorithm. Make sure the source code is updated with your specific Micro-manager config file and the paths of the model, anchors, and class are updated.

All preprocessing, training, validation, and evaluation of the neural networks was done in Python, using the following software packages:

  • Python v3.6.8
  • Tensorflow-gpu v1.14 (can be regular tensorflow-v1.14 if running on CPU)
  • Keras v2.1.5
  • Numpy v1.19.0
  • Matplotlib v3.2.2
  • Opencv-python v4.3.0.36
  • Lxml v4.5.2
  • Pillow v7.2.0
  • Scikit-image v0.16.2
  • Dippykit v3.0.0
  • pymmcore 10.0.0.1

The network parameters used include:

  1. Optimizer: ”Adam”
  2. Learning rate: 0.001
  3. Batch size: 8