Marco Fontani joins the Forensic Focus podcast to discuss Amped Software’s latest AI-powered tool, DeepPlate. DeepPlate is a deep learning system designed to read license plates affected by common issues introduced by surveillance systems, making it a valuable asset for forensic video analysts.
Marco provides an in-depth overview of DeepPlate, explaining how it currently supports eight countries (France, Germany, Italy, the Netherlands, Spain, Sweden, the UK, and the United States) and how it uses separate models for each country to account for varying license plate formats. He also discusses the synthetic data generation process used for training the models, ensuring privacy and mitigating bias.
The conversation delves into the limitations and considerations when using AI-based tools in a forensic workflow, with Marco emphasizing the importance of bias mitigation techniques and proper interpretation of confidence scores. He stresses that DeepPlate is an investigative tool designed to provide a second opinion, rather than a tool for court admissibility.
00:00 – What is DeepPlate and what is its purpose?
02:45 – How many countries does DeepPlate currently support?
09:30 – What are the challenges of diplomatic and personalised number plates?
11:30 – How would a forensic video analyst use DeepPlate as part of their workflow?
13:40 – Can DeepPlate be used in Court?
16:30 – What is Amped Software’s position on AI?
19:00 – What will be the impact of updated models on results?
20:50 – What kind of data is used to train the system?
23:30 – How has the system been tested?
26:15 – What are the access conditions for using DeepPlate?
28:05 – Does Amped Software retain user data for DeepPlate and if yes for how long?
29:45 – Is it possible to buy additional seats for your license?
31:00 – What are the limitations and considerations when using AI-based tools like DeepPlate for forensic and investigative purposes?
35:20 – Does Amped Software plan to use AI in other areas?
Show Notes
Washington State Judge Blocks Use Of AI-Enhanced Video As Evidence In Possible First-Of-Its-Kind-Ruling - https://www.linkedin.com/feed/update/urn:li:activity:7181308834370494464/
Neural Network for Denoising and Reading Degraded License Plates - https://link.springer.com/chapter/10.1007/978-3-030-68780-9_39
Introducing DeepPlate, Amped’s Investigative Tool for AI-Powered License Plate Reading - https://blog.ampedsoftware.com/2024/02/28/introducing-deepplate-ampeds-investigative-tool-for-ai-powered-license-plate-reading