Using Vector Search

Using Vector Search to Find Incidents in Video

Vector search helps you find specific incidents in video—like "slip and fall" or "person removing goggles"—by understanding the meaning behind your search, not just specific words. It analyzes visual patterns in the footage to locate events that match your description.

To use it, type a natural-language description into the search box. For example:

• slip and fall in a hallway
• a Nissan sedan with 5FG in its license plate
• people removing their goggles

Vector search is especially useful during investigations, as it reduces false negatives—helping you identify relevant video events that traditional keyword searches might miss.

Choose Vector as your Search Mode in settings menu in the upper right of the search box, or add tag:agent to your query.

Vector Search Match Percentile

In the Search Panel, you can also choose to see only the video events that match the highest with your query by specifying the Match Percentile. For example, by choosing the 5th percentile, Camio will show only the Vector Search results that score in the top 5th percentile of all matching results.

How Vector Search Works (for the Technically Curious)

Unlike keyword search, which looks for exact word matches in text, vector search uses a completely different approach based on meaning.

Here’s how it works in simple terms:

When you enter a search like "person removing goggles", or license plate "5FAW123", the system converts your entire phrase into a numeric representation called an embedding. It does the same for the description of the video content, which has already been analyzed by Camio Visual Agents and translated into embeddings that represent what’s visually happening in each event.

Instead of matching words, vector search compares the embedding of your query text to the embeddings of videos, measuring how close they are in meaning—this is called distance scoring. The closer the embeddings are in this abstract space, the more similar the meanings are. So even if the exact words don’t appear in the video description, you can still find relevant events.

This method allows vector search to surface results that traditional keyword search might miss, especially when the phrases used in the video description don't match your search query terms exactly.