Safety & Security

Synthetic data has the potential to revolutionise the field of safety and security by enabling more accurate and effective machine learning algorithms for a wide range of applications. Our technology ranges from anomalous object identification to incident detection to event recognition. Our Chameleon platform working to ensure safety and security is a supportable reality.

Gun Detection

Use Cases

  • Gun identification
  • Incident detection
  • Event recognition, often as important as detecting the object
  • Person tracking / re-identification

Chameleon Benefits

  • Taking guns into a public environment is very problematic and risks causing the public to panic. Creating the scene and scenario in Chameleon removes all of these risks
  • Once a scenario has been created in Chameleon it can be repeated multiple times to support varying lighting and weather conditions
  • Person reidentification requires large variations in human appearances to ensure the network is not over fitting to a particular type of person. Configurable actors makes this easy in Chameleon
  • Transferring the use case to a different location/place is easy in Chameleon with all assets being reusable

Gun Detection - Datapack Examples

  • Invariance Training Datapacks (the object in the scene remains constant) are created to ensure the AI network can identify the object of interest in multiple environments and from false objects
  • Then the object of interest is placed in an environment similar to where real world detection will occur to test detection in context

FAQs

Knife Crime

Use Cases

  • Object detection
  • Incident detection e.g. knife crime
  • Often the event is as important as detecting the object, so video clips of an incident are important with beginning and ending of the event clearly identified
  • Person tracking / re-identification

Chameleon Benefits

  • Creating these sort of incidents in the real world are complex and can lead to public panic. This is avoided by creating the scenes and scenarios in Chameleon.
  • Once a scenario has been created in Chameleon it can be repeated multiple times to support varying lighting, time of day and weather conditions
  • Orientation of cameras in smart city varies significantly. Chameleon can easily map real world orientations into the virtual environment and synchronise cameras across broad areas to create AI training data for person tracking and re-identification

Satellite Surveillance

Use Cases

  • New satellite use cases can be prototyped with the platform ahead of satellite deployment
  • Object identification
  • Cloud cover detection
  • Ground disturbances

Chameleon Benefits

  • Chameleon’s camera interface can be used to match a camera pipeline, an efficient way of prototyping new satellite use cases to test customer interest ahead of a launch
  • Continuous zoom is possible from above clouds to ground level with continuous high-quality rendering at all stages
  • Digital twin technology supporting FBX allows mapping of real-world areas into Chameleon

Vehicle Tracking

Use Cases

  • Vehicle of interest
  • Dangerous/damaged vehicle
  • Stolen vehicle
  • Emission zone payments

Chameleon Benefits

  • Often the number plate of a vehicle is not visible so tracking the colour, car type, accessories (roofrack, bike rack etc) becomes important
  • Synthetic (invariant) images can be used to train the network to recognize body shapes from multiple angles
  • Multiple car colours are required to test the network
  • City environments vary significantly so a broad range of scenes available from Mindtech enable a good diversity of backgrounds around the car
  • Day/night/poor weather conditions can be easily generated

Number Plate Detection and Reading

Use Cases

  • Car parking
  • Toll road payments
  • Traffic violation events
  • Insurance verification

Chameleon Benefits

  • Orientation of cameras in smart city varies significantly. In Chameleon its easy to map any camera type and orientation into a scene
  • Number plates from foreign countries can be created to expand the robustness of the network
  • Day/night/poor weather conditions can be easily generated
  • Occlusions of number plate can be created to test the AI network against

Per character segmentation of number plate is available

Enquire

If you would like to discuss how our Safety & Security Data Packs could work for your organisation, please get in touch.