BACKGROUND

CANSOFCOM personnel are tasked and trained to conduct a wide range of dangerous missions on behalf of the Government of Canada. One of these high risk missions is Hostage Rescue.

One of the most important aspects of mission success in a Hostage Rescue operation is accurate and reliable intelligence; principally gathered from observation. To ensure mission success, while reducing collateral damage, CANSOFCOM personnel continuously observe locations to determine patterns – an extremely time consuming and monotonous process.

SCENARIO

A senior Canadian Government Official has been kidnapped and is being held hostage in a [House/Office/Warehouse/Farm]. An incident response team has been dispatched comprised of both local and federal police authorities. They have each deployed drones with various camera technologies (high resolution, electro-optics, infra-red, etc.) that provide a live/near-real time feed to the analysts on the ground.

Due to certain complexities on the ground, law enforcement agencies have requested the support of the Canadian Armed Forces, and CANSOFCOM is called upon to conduct a hostage rescue operation.

The CANSOFCOM analysts have an important job of determining historical information of the hostage situation from previous footage from local security forces, in addition to live on-going footage of various drones. The analysts need to identify important key indicators (as described in the Challenge, below) that will help ensure a successful rescue operation by CANSOFCOM operators.

CHALLENGE

CANSOFCOM seeks to automate the persistent monitoring process to discern patterns of personnel (both threat actors and innocent civilians), vehicles, distinguishing features (physical, pattern, movement, etc.), and to develop normalcy in patterns, which can be used to isolate deviations in support of a hostage rescue operation.

The principle objective of automating the process is to reduce the person power required for the analysis of surveillance footage; thus allowing CANSOFCOM personnel more time and space to perform more pertinent tasks related to the operation.

PRIZES

A unique one of a kind CANSOFCOM experience, featuring:

• One-on-One discussions with CANSOFCOM Operators
• Hands-on with CANSOFCOM drone technology
• Student employment opportunity with CANSOFCOM
• Custom Shadow box with CANSOFCOM memorabilia
• CANSOFCOM and Canadian Armed Forces SWAG

CANSOFCOM GETTING STARTED WITH COMPUTER VISION WORKSHOP

https://docs.google.com/presentation/d/e/2PACX-1vQJGBsJs2rkeTEQaiCXy–M5E3CetLxVUMHqkJ0aToVMEfcPnXXSYeCzDEr0F25xkN2L1MV5rtKitAW/pub?start=false&loop=false&delayms=3000

VIDEO FOOTAGE MATERIALS

Video footage for the purposes of this CANSOFCOM hackathon challenge can be sourced from VIRAT. The VIRAT Video Dataset collection work is supported by Defense Advanced Research Projects Agency (DARPA).

• The dataset is designed to be realistic, natural and challenging for video surveillance domains in terms of its resolution, background clutter, diversity in scenes, and human activity/event categories than existing action recognition datasets.

Please visit http://www.viratdata.org/ to learn more.

• To browse and download content, visit https://data.kitware.com/
  • To use an API to access the content, read their documentation at https://data.kitware.com/api/v1

IDEAS FOR CONSIDERATION/INSPIRATION

• What kind of indicators and warnings exist to predict normalcy in patterns? (physical, pattern, movement, etc.)
  • Indexing identified content and being able to view the results in a grid/matrix/graph format
• Utilize Artificial Intelligence and Machine Learning tools to establish baseline and content identification such as:
  • TensorFlow
  • OpenCV
  • Object detection models such as YOLO, SSD or Faster_RCNN

• What other new and growing technologies can you use to interpret and manipulate large data sets?

• Can you create automated workflows that will reduce need for human analysis?
  • Can the user go back and replay a specific segment of video and verify the content analytics?
  • Can the user compare different video footage including metadata and what content has been already analyzed?
• Easily scalable and deployable web frameworks for common/shared user access
  • Produce visually appealing and meaningful interfaces that are user friendly to use/configure
  • Mapping the route of identified people/vehicles on a Map

SOLUTION

So what does solving this challenge accomplish?

Quite simply, solving this challenge will assist in one aspect of understanding the risk associated with a CANSOFCOM Hostage Rescue situation and thus increasing the likelihood of operational success and saving lives.

Solving this challenge will not just help CANSOFCOM in a Hostage Rescue situation but has multi-dimensional applications across many military and security fields.

Solving this challenge makes a real life difference to Canadians.

WE WILL FIND A WAY