View Project
Vidsy.ai is a tool for indexing, searching and analyzing visual data. It uses an LLM to convert text searches into structured queries that are run against the results of various CNNs.
Key features:
real-time camera analysis: multiple cameras can be analyzed in sync
record video on multiple tracks in sync
re-analyze recordings
search through local images and videos
share your findings with friends and colleagues
As this release is still a beta, there are some limitations:
only supports a local file system based db, for smaller datasets
the number of included models currently used for analysis is limited and is only related to faces.
LLM usage is currently limited to 10 queries a day (a previously retrieved query can be re-used unlimited)
max nr of bytes shared is limited at 20GB a month
probably still a few bugs
What's coming next:
more models for object search and such
stage 1 will be the more conventional model types (most likely Resnet based)
in a second stage I'm planning to work out another idea round CNNs in the hopes to make it easier to teach the system how to recognize things.
support for external db systems (mongodb), which will also allow multiuser use-cases (share indexes)
flesh out some more features that make things more convenient.
cloud support for data processing so that things can scale with large camera pools.
Built with
