In Broadcast - NAB SHOW 2018 - 74

74

www.inbroadcast.com | NAB Show 7-12 April 2018 | Product Preview

Machine Learning Logs
into Broadcast Content
While Cloud has helped traditional TV networks expand faster,
processing media content is still largely a manual process
that has restrained further scaling-up...

I

n 2017, Netflix users
streamed 69,444 hours
of video per minute. An
infographic released by
DOMO provides thoughtprovoking statistics about
how data and video creation
is bursting at its seam, writes
Srinivasan KA, Co-Founder, Amagi.
Today TV networks have to use
manual services for routine media
processes such as creating
segments, or inserting
ad markers in content
for monetization. In live
events, it is imperative to
create event highlights
for VoD deliveries in short
turnaround time. This is
where machine learningaugmented
content
preparation will be a game
changer.

Factory scaling of
segment creation
Linear
broadcast
model is highly reliant on
sophisticated processing of
video for ad break points
identification, credits, color
bars and blacks. As video
consumption increases, a machine
learning-based system such as Amagi
TORNADO are intelligently built

TORNADO UI

to learn from the inputs of human
content preparation specialists and
prepare segments for playout with
rapid turnaround time. This will
not only help TV networks scale up
operations faster, but also help in
improving ROI as more content can be
processed using fewer resources.
Near real-time live to VoD
Today, even the biggest of live
sports events are fast adapting to
emergence of on-demand video.
Discovery Networks, for example,
revealed a combined cross-device
approach to measure viewership
for the recently concluded Winter
Olympics 2018. However, effectiveness
of similar strategies would largely
depend on customer loyalty across
devices, and ability of TV networks to
weave a uniform live and on-demand
experience. Machine learning can help

Osprey Video at NAB Show

Osprey Video is showing new Raptor 3G and 12G capture cards and
Talon range of hardware encoders/decoders on Booth SU8107...

O

sprey Video is introducing the
Raptor I/O family, the newest
generation of Osprey's PCIe cards.
Raptor solutions range from 3G-SDI
to single-link 12G-SDI cards that have
one to eight inputs and HDMI 1.4 and
2.0 cards in single- and multi-input
configurations. This allows the user
to ingest video from SD to 4K UHD /
DCI (depending on the card.) Each SDI
input will support 16 embedded
SDI audio channels, while
the HDMI series will support
eight channels per input.
Raptor cards will fit
various hardware platforms

from M.2 to low-profile PCIe in
four and eight-lane configurations
depending on features.
Osprey Video is displaying its
hardware encoder line, which
features the Talon G2 encoder.
Designed for A/V professionals and
broadcasters looking for end-to-end
video-streaming tools for professional
workflows, the Talon G2 ingests video
from multiple formats, encodes it to
H.264, and delivers it over IP. One
of the Talon G2's most distinctive
characteristics is its front-facing LCD
touchscreen, which offers one-touch
streaming control. Once the streaming

Osprey Raptor
PCIe capture cards

profiles are set up on the back end, a
process that can be done remotely if
needed.
Both the Talon G2 and Talon G1
integrate with the Wowza Streaming
Cloud™ service, Facebook Live, and
IBM Cloud Video.
Also on display is the Talon G2
decoder, which includes features
such as multiple PIP options, SDI and
HDMI outputs, touch display for easy
configuration, and closed-caption
support. The Talon G2 decoder's
rich feature set makes it the perfect
point-to-point solution for multisite
streaming applications.
Osprey's
external
capture
devices connect directly to a user's
computer via USB 3.0 to capture
video from either 3G-SDI or HDMI
inputs. Both the 3G-SDI and HDMI
versions offer hardware scaling,
status display, and a push-button
user interface. Compatible with
Windows®, Mac®, and Linux
operating systems, the devices
are ideally suited for laptop-

bring fresh off-the-air event highlights
to mobile platforms in just minutes to
enable immediate post-match binging
on VoD. It can also reduce subscriber
churn and lend scale to VoD creation
from 100s of concurrent live channels.
Automated ad detection
Ad monetization on linear vMVPD
platforms in spite of absence of ad
markers in input streams could be the
biggest and most immediate benefit
of machine learning techniques.
A
machine
learning-based
ad
insertion platform such as Amagi
THUNDERSTORM can automatically
detect ads in linear broadcast streams
by comparing video segments with
an active ad library. With repeated
exposure to volumes of ads, the
accuracy and speed of detection can
attain efficiency levels in excess of
80 percent. This can allow vMVPD
platforms to insert ads dynamically
without having to rely on TV networks
for ad markers.
In many ways, using machine
learning is for media processing what
Cloud is for storage. By combining the
two, TV networks and content creators
can deliver a truly innovative, factoryscale model of content preparation for
broadcast.
www.amagi.com

streaming applications using the
standard UVC and UAC drivers.
Osprey Video's lineup of end-to-end
live-streaming workflow products
complement its flagship capture cards.
In addition to SDI-to-HDMI and HDMIto-SDI converters and miniconverters,
Osprey Video offers scaling converters
for 3G-SDI to HDMI and vice versa,
along with 3G-SDI fiber extenders.
The Osprey Video SDI multiviewer
has four 3G-SDI inputs, two 3G-SDI
outputs, and one HDMI output.
The multiviewer is capable of up/
downscaling, frame-rate conversion,
color-space
conversion,
and
deinterlacing, while its information
overlays depict resolution, frame
rate, channel identification, and audio
meters. Users can operate the unit via
DIP switch or Ethernet controls.
www.ospreyvideo.com

Osprey Talon G2 range
http://www.inbroadcast.com http://www.amagi.com http://www.amagi.com http://www.ospreyvideo.com http://www.ospreyvideo.com
Table of Contents for the Digital Edition of In Broadcast - NAB SHOW 2018
