Technical architecture was fun to build. Runs on a single 2GB VM + Postgres
- Ingest scripts across 468 sources of different camera formats
- Health check to ensure only working cameras are shown
- Proxy vs direct to ensure cameras work - some serve residential IPs but firewall datacenter rangers. Others are the reverse
- Image fingerprinting to capture dead cameras with static images
- For some hls the origin delivers segments slower than the video bitrate, so it never plays. I pull a single frame out with ffmpeg and show a slow-refreshing still instead of a black box. autoidentified
- Humbled by SEO- google seemingly flagged as scaled non-quality content. Should’ve dripfed. Oh well
Proud of the build. happy to answer questions.
Technical architecture was fun to build. Runs on a single 2GB VM + Postgres - Ingest scripts across 468 sources of different camera formats - Health check to ensure only working cameras are shown - Proxy vs direct to ensure cameras work - some serve residential IPs but firewall datacenter rangers. Others are the reverse - Image fingerprinting to capture dead cameras with static images - For some hls the origin delivers segments slower than the video bitrate, so it never plays. I pull a single frame out with ffmpeg and show a slow-refreshing still instead of a black box. autoidentified - Humbled by SEO- google seemingly flagged as scaled non-quality content. Should’ve dripfed. Oh well Proud of the build. happy to answer questions.
Is there no way to report incorrectly labelled cameras? Found two that are nowhere near where they claim to be. Otherwise, super neat!
Nevermind, found it, reported the discrepency, thanks for the work!