Vattenjagen

#8 — Rian Design Museum, 2025

Vattenjagen #8 is a mechatronic installation enabling dialogue with Lake Bolmen—a prototype exploring communication between human and non-human. The work invites visitors to write in the sand along the shoreline. The lake "responds" with an image, projected onto the water's surface, creating a moment of exchange between visitor and landscape.

The installation operates at dawn and dusk, when the boundary between water and sky becomes fluid. Visitors arrive in silence, trace characters in wet sand, and wait. The response—a fleeting projection—dissolves back into the lake's surface, as if the water itself is thinking, remembering, imagining.

This work continues Studio Alight's investigation into liminal spaces: thresholds where human perception meets machine vision, where language meets landscape, where the seen meets the sensed. The lake becomes a mirror, a memory, a participant in dialogue.

Location: Rian Design Museum, Falkenberg

Year: 2025

Medium: Mechatronic installation, projection mapping, computer vision

Exhibition: Part of ongoing Vattenjagen series

Vattenjagen

#8 — Technical Documentation

System Architecture

STEREO CAMERA ARRAY → YOLO OBJECT DETECTION → IMAGE CLASSIFICATION ↓ ↓ ↓ SAND TRACKING HUMAN DETECTION CONTENT SELECT ↓ ↓ ↓ TRIGGER SIGNAL → PROJECTION MAPPING → LAKE SURFACE OUTPUT

Hardware

• Camera Array: One 4K camera with IR illumination for low-light conditions

• Processing Unit: NVIDIA Jetson AGX Orin for edge computing. Runs YOLOv5 for real-time human detection at 30fps.

• Projection System: 12,000 lumen laser projector. Calibrated to water surface geometry.

Software Stack

• Object Detection: YOLOv5 trained on COCO dataset, fine-tuned for shoreline human silhouettes

• Classification: ImageNet-based classifier (ResNet50) for style and content matching

• Projection Mapping: OpenCV calibration with water surface distortion compensation

Data Flow

1. Stereo cameras capture shoreline at 30fps 2. YOLO detects human presence and tracks motion path 3. Writing gesture triggers: motion stops + sand contact detected 4. System captures written character/image from bounding box 5. Character classified (letter? symbol? abstract shape?) 6. Content database queried for semantic match 7. Selected image warped for water surface projection 8. Laser projector outputs for 3-5 seconds 9. Fade to black, system resets

References

Based on prior work: [Self] Confidence (2019) — neural network classification research

GitHub: github.com/studioalight/vattenjagen

Paper: "Mechatronic Dialogues: Water as Interface" (in review)