A private pan/tilt camera that patrols the room and confirms sightings with a custom-trained second model.
“Self-Hosted Pet Camera That Searches the House for Your Dog” is an advanced Security, Science & Hardware AI agent workflow from Any. It connects Other or custom in 4 steps. Setup takes about Not stated and running costs are about Not stated.
Off-the-shelf pet cameras push your living room video through a vendor cloud and only show what the lens happens to point at, so you still have to steer them manually to find an animal. Generic object detectors also misfire on plush toys and furniture, making automated alerts unreliable. This build keeps footage on your own server, reaches the camera over a private tunnel instead of exposing the home network, and adds a verification layer trained on your own pet so only real detections count.
Take a low-cost pan/tilt camera and install open firmware on it so you control the software stack. This removes the vendor app and cloud dependency and gives you direct access to the motor controls and video stream.
Have the camera initiate an outbound connection to a small rented server over a VPN tunnel. Because the camera dials out, no inbound ports are opened on the home router, and the server can relay the video feed to your phone.
Use a lightweight detector to find candidate animals in each frame, then pass every bounding box through a second classifier fine-tuned on crops from your own camera. Set the acceptance threshold from the weakest reference photo of your pet instead of a fixed value, so the verifier rejects plush toys and lookalikes.
Let the server keep a rolling buffer of recent footage and drive the pan/tilt motors to sweep the room when you ask where the pet is, rather than requiring manual joystick control. Stream the result to your phone over WebRTC.
You are helping me plan a self-hosted pet camera. Outline the components: an open-firmware pan/tilt camera, an outbound VPN tunnel to a small VPS, a WebRTC relay to a mobile client, a rolling recording buffer, and a two-stage detector. For the detector, describe how to collect crops from [MY CAMERA FEEDS], fine-tune a small image classifier on them, and choose an acceptance threshold from the weakest reference photo of [MY PET] rather than a fixed cutoff. List the config values I need to fill in and the failure modes to test, such as plush toys and poor lighting. Do not invent prices or hosting providers.
Write pseudocode for a two-stage pet detection service. Stage one runs a lightweight object detector on frames from [CAMERA STREAM] and emits bounding boxes. Stage two crops each box and scores it with a classifier fine-tuned on images of [MY PET], accepting only scores above a threshold derived from the lowest-scoring reference photo. Include how to log rejected boxes for later review, how to trigger a pan/tilt sweep when no confirmed detection appears, and how to keep a rolling buffer of recent footage. Keep it framework-agnostic and note where I must supply model paths and credentials.
A private pan/tilt camera that patrols the room and confirms sightings with a custom-trained second model.
You need Other or custom.
Plan for about Not stated. It is rated advanced. The guide walks through 4 steps, starting with “Replace the camera firmware”.
Expect about Not stated in running costs (Under $25/mo range).
Off-the-shelf pet cameras push your living room video through a vendor cloud and only show what the lens happens to point at, so you still have to steer them manually to find an animal. Generic object detectors also misfire on plush toys and furniture, making automated alerts unreliable. This build keeps footage on your own server, reaches the camera over a private tunnel instead of exposing the home network, and adds a verification layer trained on your own pet so only real detections count.
An assistant on top of Gmail and Calendar detects recurring meetings, repeated email types and bills, then asks whether to prep or schedule them.
Original post ↗Users' AI agents phone hotels after making a reservation to confirm it was booked correctly, per a hotel worker complaining about the calls.
Original post ↗An agent researches keywords, publishes one SEO article daily and refreshes articles every two weeks using Search Console and Ahrefs data.
Original post ↗Built this workflow? Show it is featured on AgentUseHub with a live upvote count that links back here. About badges
