qr-tiling

QR Tiling Detection

This application demonstrates high-resolution QR code detection using dynamic image tiling, running fully on-device with a live frontend for configuring the tiling grid.

It runs QR detection on the DepthAI backend with dynamic tiling, and exposes controls in the UI for:

• Configuring grid size (rows and columns)
• Adjusting tile overlap percentage
• Merging tiles into custom regions via interactive grid editor
• Enabling global detection (full-frame pass)
• Real-time tile grid visualization overlay

Note: RVC4 standalone mode only.

---

Features

• Dynamic tiling for high-resolution detection

  • Split high-resolution frames into smaller tiles for neural network inference
  • Configurable grid size up to 8×8 tiles
  • Adjustable overlap between tiles to catch QR codes on boundaries

• QR code decoding

  • Detects QR codes using neural network
  • Decodes QR content using pyzbar
  • Displays decoded URLs/text as annotations

• Custom tile merging

  • Interactive grid editor in the UI
  • Click to select cells, click adjacent cells to merge
  • Create custom tile regions for optimized detection areas

• Global detection mode

  • Optional full-frame inference pass
  • Catches large QR codes that span multiple tiles

• Real-time visualization

  • Tile grid overlay on video stream
  • Color-coded tiles with blended overlap regions
  • Live tile count display

• Runtime configuration

  • Update tiling parameters without restarting
  • UI state synchronized with backend
  • Configuration persists across UI refreshes

• Adaptive FPS control

  • Monitors pipeline backpressure (BLOCKED input queues) and adjusts FPS to prevent overload
  • Estimates safe FPS from node timing data when tile count increases
  • Gradually ramps FPS back up after tile count decreases

---

Architecture

---

Usage

Running this example requires a Luxonis RVC4 device connected to your computer. Refer to the documentation to set up your device if you haven't already.

---

Standalone Mode (RVC4)

To run the example in standalone mode, first install the oakctl tool using the instructions here.

The app can then be run with:

oakctl connect <DEVICE_IP>
oakctl app run .

Once the app is built and running you can access the DepthAI Viewer locally by opening https://<OAK4_IP>:8082/ in your browser (the exact URL will be shown in the terminal output).

Remote Access

You can upload the oakapp to Luxonis Hub via oakctl and then remotely open the App UI via the App detail page.

---

Configuration

Note: FPS may drop briefly after increasing tile count (to prevent overload) and will gradually ramp back up after decreasing the tile count.

Tiling Parameters

Parameter	Range	Default	Description
rows	1-8	2	Number of tile rows
cols	1-8	2	Number of tile columns
overlap	0.0-0.99	0.2	Overlap percentage between adjacent tiles
global_detection	bool	false	Include full-frame inference pass
grid_matrix	2D array	null	Custom tile merging configuration

Resolution Settings

Defined in main.py:

TILING_SIZE = (3840, 2160)
OUT_SIZE = (1920, 1080)