Hack-a-Bot
AI-powered porch defense with person detection and precision water targeting
The kit includes the servo turret, solenoid valve, nozzle, and detection electronics. You supply:
The Full Send Kit includes all electronics and hardware. You supply:
Mount on your porch, connect a garden hose, and let AI handle the rest.
Every Zero Shot unit is hand-assembled, firmware-loaded, sensor-calibrated, and QA tested before shipping. Premium boxed packaging included.
Upgrade any porch camera or doorbell with AI-powered package defense. Person detection triggers recording and alert. If a person enters the package zone and picks up a package, a servo-controlled water nozzle delivers a precision blast. This is machine learning meets practical deterrent — non-lethal, legal (on your property), and deeply satisfying.
The system uses a Raspberry Pi 4 running MobileNet SSD for real-time person detection, a pan servo for target tracking, and a solenoid-controlled water valve for delivery. All controlled via a phone app with arm/disarm scheduling, live video feed, and event logging. Water only — no projectiles, no pepper spray, no legal gray areas.
AI Demonstration: Live detection accuracy test. System must distinguish between person, animal, and package with at least 90% accuracy across 20 test scenarios.
Utility Challenge: Real-world reliability over 7 days. Judges score on false positive rate, response time, and system uptime.
Freestyle: Innovation and build quality. Bonus points for advanced features like facial recognition, multi-zone defense, or integration with home automation.
Mount a Pi Camera or USB webcam with a clear view of your porch and package drop zone. Position it high enough to avoid tampering but low enough for good facial detail. Wide-angle lens (120-160 degrees) covers more area but reduces resolution at distance.
Test the view before permanent mounting. Mark the package zone on the ground (tape or chalk). Verify the camera can see the entire zone plus approach paths. Lighting matters — install under an overhang or add an IR illuminator for night operation.
MobileNet SSD runs on the Pi 4 at 10-15 FPS — fast enough for real-time detection. The model detects people, packages, animals, and vehicles. You care about people and packages. Everything else is ignored.
Define a region of interest (ROI) where packages typically sit. When a person enters the frame AND enters the package ROI, start tracking. If the person picks up the package (package bounding box disappears while person is still in ROI), trigger the water system.
// Detection logic (simplified)
if (person_in_frame && person_in_package_zone) {
track_target(person_bbox);
// ... full detection pipeline in project files
Pan servo mounts on a bracket attached to your camera housing or porch rail. The water nozzle attaches to the servo arm. When a person is detected, calculate the horizontal position of their bounding box and steer the servo to aim at them.
Servo range: 0-180 degrees. Map camera frame X-coordinate to servo angle. The kit includes a servo with metal gears (critical — plastic gears strip under water pressure). Calibrate the mapping so center-frame equals center-servo-position.
// Servo aiming (simplified) target_angle = map(person_bbox.center_x, 0, frame_width, 0, 180); pan_servo.write(target_angle); // ... full tracking controller in project files
12V solenoid valve controls water flow from a garden hose. The valve sits inline between the hose and the nozzle. When the Pi sends a signal, a MOSFET relay activates the valve and water flows for 1-2 seconds (configurable burst duration).
Nozzle choice matters. Adjustable spray nozzles let you tune from mist to jet. For competition, use a focused stream — more range, less collateral spray. Mount the nozzle on the servo arm with zip ties or a 3D-printed bracket.
When a person is detected, the system saves a video clip and sends a push notification to your phone. Use Pushover, Telegram bot, or email with embedded snapshot. Include timestamp, detection confidence, and a link to the live feed.
If the water system fires, log the event with before/after video clips. This is critical for legal documentation if someone claims excessive force or property damage. Water is generally legal, but documentation protects you.
Phone app or web interface lets you arm/disarm the system. Add scheduling — disarm during expected delivery windows (10am-2pm), arm overnight and during known risk times. Include a "visitor mode" that disables the water but still records and alerts.
Physical override: a keypad or RFID tag near the door lets trusted visitors disarm the system. Delivery drivers with your access code won't get sprayed. Porch pirates won't have the code.
Dogs, cats, shadows, and reflections can trigger person detection. Use a PIR motion sensor (included in kit) as a pre-filter — only run the AI detection when motion is detected. Saves CPU and battery, reduces false positives from static objects.
Confidence threshold: MobileNet SSD outputs a confidence score for each detection. Set a minimum threshold (e.g., 70%) to filter out low-confidence detections. Adjust based on your environment — more false positives means raise the threshold, more missed detections means lower it.
Complete source code, wiring diagrams, and documentation. Enter your kit code to unlock.
Don't have a code? Get the kit — code included in every box.