We are transforming the way shoeprints from crime scenes are utilized

We live in a world filled with cameras—on streets, inpublic places, residential areas, and storefronts. When a criminal case occurs,these cameras act as 'eyes', providing valuable assistance for investigations.

However, in cases of theft, crime scenes typically lack cameras. As offenders become more adept at counter-surveillance, it is becoming increasingly difficult to find fingerprints or DNA evidence. Yet, more than 55% of burglary scenes still retain shoeprint evidence. How to effectively use this shoeprint evidence to aid investigators in solving cases has become a focal point in forensic research.

If we could combine the widespread cameras in cities with shoeprint evidence left at crime scenes, identifying suspects would become much easier. After years of research into shoeprint-matching AI algorithms and video-analysis AI algorithms, a complete solution is now available. This solution allows investigators to identify the type of shoe from the shoeprint left at the scene, then use the shoe's characteristics to search for individuals wearing those shoes in video footage, ultimately narrowing down suspects. This solution has already been piloted in several forensic labs and is continually being validated and refined.

Searching for shoes from scene shoeprints can yield many similar shoe models, and manually identifying people wearing those specific models in video footage is a challenging task. Here’s an example case to illustratethis: we have a 60-minute, 1080p video featuring 591 individuals, and we need to search for three specific shoe models within the footage.

One person watching the video at 2x speed (generally, videos are reviewed ny people at speeds of up to 8x fastest) would need 30 minutes to view it, plus an additional 6 minutes to match similar shoes in the footage, totaling 36 minutes. The constant focus on the video is also very taxing on the eyes.

In comparison, a standalone Everspry Video Tracking System (EverVTS) can analyze video at 20 times the normal speed, taking 4 minutes to load and automatically parse an hour-long video, and only 1 additional minute to match with the three shoe types. Reviewing the match results takes about 4 minutes, totaling approximately 8 minutes overall.

While the EverVTS addresses the time-consuming nature of reviewing footage, there are still significant limitations that remain unresolved, such as low video resolution and poor camera angles. As technology advances, we believe that the application of shoeprint evidence will provide substantial assistance to crime-fighting efforts worldwide in the future.

We are currently seeking 10 test users globally. If you have video analysis needs, you can send us the video and shoeprint images from the scene, and we will provide a free analysis service. We guarantee confidentiality for all data, and after delivering the comparison results, we will permanently delete all materials.

‍