PiRATeMC: A highly flexible, scalable, and affordable system for obtaining high quality video

recordings for behavioral neuroscience.

Authors: Samuel W. Centanni, PhD

1

, Alexander C.W. Smith, PhD

2

*

1

Department of Molecular Physiology & Biophysics, Vanderbilt University, Nashville, TN

2

Nash Family Department of Neuroscience, Icahn School of Medicine at Mount Sinai, New York, NY

*Corresponding Author:

Abstract

With the recent development and rapidly accelerating adoption of machine-learning based

rodent behavioral tracking tools such as DeepLabCut, there is an unmet need for a method of acquiring

video data that is scalable, flexible, and affordable. Many experimenters use webcams, GoPros, or

other commercially available cameras that are not only relatively expensive, but offer very little flexibility

over recording parameters. These cameras are not ideal for recording many types of behavioral

experiments, and can lead to suboptimal video quality. Furthermore when using relatively affordable

commercially available products, it is a challenge, if not impossible, to synchronize multiple cameras

with each other, or to interface with third-party equipment (for example, receiving a simple trigger to

synchronously controlled in clusters via a self-contained network to record a large number of

simultaneous behavioral sessions without burdening institutional network infrastructure. Furthermore,

the Raspberry Pi is an excellent platform for novice and inexperienced programmers interested in using

an open-source recording system, with a large online community that is very active in developing novel

open-source tools. It easily interfaces with Arduinos and other microcontrollers, allowing simple

synchronization and interfacing of video recording with nearly any behavioral equipment using GPIO

pins to send or receive 3.3V or 5V (TTL) signals, I2C, or serial communication.

.CC-BY-NC-ND 4.0 International licenseavailable under a

was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

INTRODUCTION

Preclinical studies have long relied on traditional predetermined activity patterns to assess

specific groups or subgroups within experiments. Moreover, manual behavioral scoring is highly

vulnerable to human error, biases, and low inter-rater reliability. Using unbiased pose estimation

techniques with open-source machine learning-based software such as DeepLabCut

mapping or clustering analysis with packages such as B-SOID

behavioral neuroscience into a new era of behavioral analysis and categorization. These techniques

are able to segment behaviors in an unbiased way, and eliminate inconsistencies due to human error

and inter-rater variability. To maximize the potential for these and other programs, experimenters must

be able to capture high resolution videos. Moreover, the ability to interface recording equipment with

single-board computer and an accompanying 8-megapixel (MP) Camera Board (Sony IMX219 CMOS

sensor). These cameras can record high quality video under either infrared (IR) or white light. Moreover,

they offer far more flexibility in recording parameters than most commercially available camera

systems. The user can manually set recording parameters like frame size, frame rate (up to 120FPS),

white-balance (crucial for high-quality IR videos), brightness/contrast, ISO, saturation, bitrate, and

many more (details in Table 2 and Supplemental Table 2). Finally, the PiRATeMC system can easily

be controlled remotely via ssh (remote secure shell) over a local area network (LAN), and clusters of

cameras can be controlled synchronously with millisecond precision using ClusterSSH (easily installed

via the advanced packaging tools (apt) in Linux or Homebrew in MacOS), allowing either multi-angle

recording of single subjects (e.g. for 3D-DeepLabCut

), or an easy method of recording a large number

.CC-BY-NC-ND 4.0 International licenseavailable under a

was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

on a remote local network (LAN) host. The goal of this paper is to provide an easily obtainable recording

system that is highly flexible in recording parameters, can interface with numerous real-time equipment

METHODS

Note that any text highlighted in grey below represents inputs or outputs from the Linux terminal, and is case-sensitive.

Camera Assembly

Raspberry Pis (RPis) are single-board computers that run a full Linux operating system

(Raspbian), and thus offer greater computational capabilities than other single-board computing

devices such as Arduinos. For a simple recording setup, we recommend using the RPi 3 Model B+, as

this model is the last to have a standard HDMI output port and micro-USB charger. RPi 4 and newer

can certainly be used, however newer models only have micro-HDMI output, and a USB-C power input,

which likely necessitates purchase of additional adapters. There are two options for 8MP camera

is only 6” long, so while a longer Flex Cable is not absolutely necessary, we consider a longer flex cable

necessary equipment, and the price is very low.

.CC-BY-NC-ND 4.0 International licenseavailable under a

was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

Table 1. Parts List. This table contains the minimum essential parts to get started with one camera.

When inserting flex cables, the silver leads of the cable always face away from the slider clamp that

secures the cable in place, on both the camera and the RPi. Instructions and example photos for

attaching the camera to the RPi shown in Figure 1.

Remote Controller Hardware Configuration

While RPis can be connected to a keyboard and monitor and operated as standalone computers,

the most convenient way to control the cameras is via a remote secure shell (SSH) connection. This

allows the RPi to be controlled from a remote computer (referred to here as the ‘remote controller’)

connected to the same network as the RPi, and the remote controller can also serve as a data sink

where videos are automatically transferred after recording. Because the RPi operating system is a

Linux distribution (Raspbian), it is most convenient if the remote controller is also running a UNIX-based

.CC-BY-NC-ND 4.0 International licenseavailable under a

was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

departments prohibit use of network switches, the remote controller should have at least a dual-port

network adapter card, as transmitting video data over WiFi is orders of magnitude slower. This will allow

architecture is shown in Figure 2. A network switch can then be connected to the second port, allowing

the remote controller to assign and manage IP addresses to any RPis connected to the switch, without

the RPis having access to the main institutional network. Thus, PiRATeMC does not pose any security

risk at hospitals/university medical centers

and other institutions where security is a

significant concern. Detailed instructions

for configuring this network are provided in

supplemental materials, and an

.CC-BY-NC-ND 4.0 International licenseavailable under a

was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

https://www.raspberrypi.org/software/, or by typing: sudo apt install rpi-imager into a Linux terminal. We

.CC-BY-NC-ND 4.0 International licenseavailable under a

was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

.CC-BY-NC-ND 4.0 International licenseavailable under a

was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

move files to/from the RPi and any computer connected to the same network with the scp command,

for details input man scp into the terminal.

the RPi through an SSH connection, you are nearly ready to begin recording or streaming videos. The

first time you power on each RPi, the first command entered should be sudo raspi-config. This will bring

you to a GUI configuration manager, where it is advised to change the password, the hostname (if you

did not change it in the rootfs while flashing the SD card), and the location/time zone settings. If you

are based outside of the USA, it may be necessary to change the keyboard layout as well. There are

several methods for recording or streaming videos.

.CC-BY-NC-ND 4.0 International licenseavailable under a

was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

Recording via recordVideo.sh script:

The recordVideo.sh script is the most powerful method for recording high-quality videos, as it

location set in the $REMOTEVIDPATH

variable above. Please note that videos

will also be saved locally on the RPi,

and should therefore be deleted

regularly to avoid running out of

storage. This is critical, as full storage

the recording. To clear all videos from

safer, but more time-consuming way is to use the rm command with the exact filename (e.g., rm

TestVideo1.mp4), or to use a shared string within video names (e.g. rm Test*.mp4). To confirm the files

are removed, use the command ls *.mp4. No files with an .mp4 extension should be present.

Table 2 shows some of the most useful recording parameters that can be manually set. To

change these parameters, edit the recordVideo.sh script with nano recordVideo.sh, and parameters

can be inserted in line 3 after the raspivid command. The impact of some of these recording parameters

on video quality is shown in Figure 4. A caveat of the minimal setup we have described here is that the

recordVideo.sh script does not easily allow simultaneous viewing of the video while it is being recorded.

This can be remedied by attaching a 2.8” touchscreen to the RPi and running the command: sudo

python3 ~/setupScripts/adafruit-pitft.py, and following the prompts. The final prompt of the setup

process will ask “Would you like the console to appear on the screen?”, and answering “No” to this will

.CC-BY-NC-ND 4.0 International licenseavailable under a

was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made