Introduction to Video Ecosystem Mind Map Video Streaming Platform Video Ad Tech Platform Video Player Platform Video Content Distribution Platform Video Analytics Platform Summary of key ideas Q & A

1. Video Ecosystem and some
ideas about video big data
Trieu Nguyen - Head of Platform at Blueseed Digital
My personal email: tantrieuf31@gmail.com

2. Agenda
1. Introduction to Video Ecosystem Mind Map
a. Video Streaming Platform
b. Video Ad Tech Platform
c. Video Player Platform
d. Video Content Distribution Platform
e. Video Analytics Platform
2. Summary of key ideas
3. Q & A

3. https://video-guide.iab.com/new-tv

4. The value of video content
1. Retain user loyalty with multimedia content
2. Builds the volume of user traffic very fast
3. Delivers best content UX to your user
4. Give your business a wider marketplace

6. (1) Video Streaming Platform

9. https://www.koeppeldirect.com/drtvblog/rise-of-livestreaming-marketing-trends-tips

13. Video Streaming Market Research Report- Global Forecast 2023
https://www.marketresearchfuture.com/reports/video-streaming-market-3150

14. (2) Video Advertising Technology Platform

17. How Video is used for Digital Market → Aha: Viral Clip

18. Omni-channel Marketing

19. Is YouTube a new TV, and compete directly with traditional TV ?

20. https://www.iab.com/wp-content/uploads/2018/04/2018_IAB_NewFronts_Video_Ad_Spend_Report.pdf

21. (3) Video Player Platform

23. http://MediaPlayer.one

24. (4) Video Content Distribution Platform

28. https://www.slideshare.net/xamat/qcon-sf-2013-machine-learning-recommender-systems-netflix-scale
In Video Content Platform,
the Recommendation Engine is key feature

29. (5) Video Big Data Platform

30. Video Analytics Platform
i. What is Video Big Data ?
ii. What is Visual Information Analytics and why ?
iii. How can we extract value from video ?
iv. How we design Video Big Data System ?

31. Video Big Data — Examples
● Netflix — “Other Movies You May Enjoy”
● YouTube — “Recommended Videos”
● Aventura security — “Cerebrus Intelligent Video Analytics”

32. What is Video Big Data ?

33. Visipedia, short for “Visual Encyclopedia,”
Visipedia, is a network of people and machines that is designed to harvest and organize visual information
and make it accessible to anyone anywhere ( https://visipedia.org )

35. https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/45530.pdf

38. Pre-trained
models
+
+
Metadata
User Interface
Distributed computing resources.
E.g. Servers with/without GPUs.
Developers / Scientists
Users / Annotators / Analysts
Deep Learning
libraries
?
How do we structure Visual Data processing?
?
?
Web, Streaming cameras
and other external sources

39. Set of images
Video / Dataset
Frame
Perform_dataset_extraction
perform_detection
TF object detection API,
CTPN, MTCNN, etc.
IndexEntries
Region
Sun
Yosemite
perform_indexing
Inception, vgg,
facenet etc.
Regions are 2D bounding boxes on a frame and
can be generated via detectors / annotators or
provided via UI, REST API or pre existing
metadata. Regions also JSON and text
metadata. And can be “Materialized” as a
separate image.
peform_analysis
Open Images tags,
CRNN text recognition,
etc.
Async tasks are underlined
Each box is a data model
Segment
Each segment begins
at an I-type Keyframe
This enables parallel
decode/processing of
video across multiple
machine in chunks.
perform_video_segmentation
perform_indexing
Inception, vgg, facenet etc.
IndexEntries stores filenames of
numpy arrays containing
features and corresponding
JSON files.
Tubes
Sun
Yosemite
Tubes are sequences of Regions.
Tubes can be used to represent
set of regions or frames or
segment for storing metadata
about “tracks”, “clips” etc.
perform_segment_decode
detect_scenes

40. REST API
+
DVAPQL
Video, image, frame,
region, etc.
metadata stored in
Relational database
Filesystem with images, videos, feature
vectors & models.
Task queues
Segmentation
workers
Indexing
workers
Nearest
Neighbors
workers
Video segmenter
/ decoder /
encoder
Detection
workers
Analysis /
Annotation
workers
Image
processing
workers
Export
Import / Ingest Scheduler for
periodic tasks
Data-centric Architecture for Video Big Data
Source:
https://www.deepvideoanalytics.com

41. Key takeaways
1. Video Streaming Market is really hot
2. Digital Video Platform is the key to
success
3. Big Data is “must-have” system for
“Video-First Business” → Video Big Data