serenity

Basically improv + mesmerize + fastplotlib for realtime analysis and visualization of calcium imaging and behavior data.

WIP

Installation

Install zmq in matlab

Build jeromq using maven on linux and use the jar file on matlab in Windows. I was not able to build on windows.

1. Install OpenJDK 8 on linux. Newer versions of Java beyond v8 do not work in matlab! basically just unpack this tar.gz file and then export the unpacked path as JAVA_HOME and set the PATH to the bin dir, and LD_LIBRARY_PATH to the lib dir: https://adoptium.net/temurin/releases/

2. Install maven and make sure it binds to java 8: https://dlcdn.apache.org/maven/maven-3/3.9.1/binaries/apache-maven-3.9.1-bin.zip

export JAVA_HOME=...
export PATH=...
export LD_LIBRARY_PATH=...
sudo apt install maven
mvn --version

Should return:

Apache Maven 3.6.3
Maven home: /usr/share/maven
Java version: 1.8.0_362, vendor: Temurin, runtime: /home/kushalk/Downloads/jdk8u362-b09/jre
Default locale: en_US, platform encoding: UTF-8
OS name: "linux", version: "5.10.0-21-amd64", arch: "amd64", family: "unix"

Make sure it says Java version: 1.8.x, 1.8 is Java 8 apparently.

3. Build jeromq

cd ~/repos
git clone https://github.com/zeromq/jeromq.git

# checkout v0.5.3
git checkout v0.5.3

# build jar
mvn package

4. Copy jar file to scanimage computer

scp kushalk@hantman-calcium:/home/kushalk/repos/jeromq/target/jeromq-0.5.3.jar ..

To test:

Start server in python on receiving computer:

import zmq

context = zmq.Context()
sub = context.socket(zmq.PULL)
sub.bind("tcp://0.0.0.0:9050")
sub.recv()

# should receive
Out[5]: b'yay'

Start client on matlab and send:

javaaddpath('C:\Users\scanimage\jeromq-0.5.3_java8.jar')
import org.zeromq.*

context = ZContext()
socket = context.createSocket(SocketType.PUSH)
socket.connect("tcp://152.19.100.28:9050")
socket.send("yay")

Benchmarking, send random frames from matlab:

t0 = clock();
for i = 1:2000
  a = int16(randi(512, 512));
  socket.send(getByteStreamFromArray(a));
end
clock() - t0

This gives ~6.4 seconds, or ~300 frames/s

l = list()
while True:
    t0 = time()
    b = sub.recv()
    l.append(time() - t0)

This results in a median delay of ~0.0044 s between frames, or ~225 frames/second, which is just below the theoretical max of 238 for a 1 gigabit connection since 125 MB / (512 512 2) = 238. 125MB = 1 gigabit.