---

name: rdkx5-app-resources description: "Access to RDK X5 /app folder resources including GPIO, multimedia, AI samples. Invoke when user wants to run embedded demos or control hardware on RDK X5."

RDK X5 App Resources Skill

This skill provides access to embedded hardware resources in /app folder on D-Robotics RDK X5 boards, including GPIO control, I2C/SPI communication, video capture, media processing, and AI inference.

⚠️ Important: Use System Python (NOT conda)

Most of these resources require system Python, NOT conda environment!

The D-Robotics board has hobot_dnn and other libraries installed in system Python (/usr/bin/python3.10), but NOT in conda environments.

Always use system Python to run these scripts:

# ✅ Correct - use system Python
/usr/bin/python3.10 /app/pydev_demo/01_basic_sample/test_resnet18.py

# ❌ Wrong - conda environment does NOT have hobot_dnn
python /app/pydev_demo/01_basic_sample/test_resnet18.py  # will fail!

If you accidentally use conda Python, you'll get errors like:

ModuleNotFoundError: No module named 'hobot_dnn'

Available Resources

1. 40pin GPIO Samples (/app/40pin_samples/)

Python-based GPIO control using Hobot.GPIO library:

File	Description
simple_out.py	Basic GPIO output (LED control)
simple_input.py	Basic GPIO input (read pin state)
simple_pwm.py	PWM output (e.g., motor speed control)
button_event.py	Button event handling
button_interrupt.py	Button interrupt handling
button_led.py	Button controlling LED
test_i2c.py	I2C communication test
test_spi.py	SPI communication test
test_serial.py	Serial/UART communication test
test_all_pins.py	All pins test

2. Character Device Demos (/app/cdev_demo/)

C-based device driver examples:

Directory	Description
v4l2/	V4L2 video capture
vio_capture/	Video input capture
vio2display/	Video to display output
vio2encoder/	Video encoding
rtsp2display/	RTSP stream to display
vps/	Video Processing System
bpu/	BPU (Brain Processing Unit) demos
decode2display/	Decode to display

3. Multimedia Samples (/app/multimedia_samples/)

C/C++ multimedia processing examples:

Directory	Description
sample_codec/	Audio/Video encoding & decoding
sample_dsp/	DSP processing
sample_gdc/	Geometric Distortion Correction
sample_gpu_2d/	2D GPU operations
sample_hbmem/	High Bandwidth Memory
sample_isp/	Image Signal Processor (camera tuning)
sample_osd/	On-Screen Display
sample_pipeline/	Media pipeline processing
sample_usb/	USB camera support
sample_vin/	Video Input
sample_vot/	Video Output
sample_vse/	Video Sync Engine
sunrise_camera/	Sunrise camera modules
vp_sensors/	Vision platform sensors
chip_base_test/	Chip baseline tests
utils/	Utility tools

4. Python AI Samples (/app/pydev_demo/)

Python-based AI model inference examples:

Directory	Description
01_basic_sample/	Basic usage
02_usb_camera_sample/	USB camera inference
03_mipi_camera_sample/	MIPI camera inference
04_segment_sample/	Image segmentation
05_web_display_camera_sample/	Web display camera
06_yolov3_sample/	YOLOv3 object detection
07_yolov5_sample/	YOLOv5 object detection
08_decode_rtsp_stream/	RTSP stream decoding
09_yolov5x_sample/	YOLOv5x object detection
10_ssd_mobilenetv1_sample/	SSD MobileNetV1 detection
11_centernet_sample/	CenterNet object detection
12_yolov5s_v6_v7_sample/	YOLOv5s v6/v7 variants

Model files are linked from /app/model/basic (or /opt/hobot/model/x5/basic).

5. ISP Tuning Tool (/app/tuning_tool/)

Image Signal Processor tuning tools:

File	Description
isp_tuning	ISP tuning executable
run_tuning.sh	Tuning launch script
tuning_cfg/	Tuning configuration files

Quick Reference

Running Python Samples - ALWAYS use system Python!

# ✅ CORRECT - Use system Python (/usr/bin/python3.10)
# These have hobot_dnn and other D-Robotics libraries

/usr/bin/python3.10 /app/40pin_samples/simple_out.py
/usr/bin/python3.10 /app/40pin_samples/button_led.py
/usr/bin/python3.10 /app/40pin_samples/test_i2c.py
/usr/bin/python3.10 /app/40pin_samples/test_spi.py
/usr/bin/python3.10 /app/40pin_samples/test_serial.py

/usr/bin/python3.10 /app/pydev_demo/01_basic_sample/test_resnet18.py
/usr/bin/python3.10 /app/pydev_demo/07_yolov5_sample/main.py
/usr/bin/python3.10 /app/pydev_demo/08_decode_rtsp_stream/main.py

Building C/C++ Demos

# Build multimedia samples
cd /app/multimedia_samples/
make

# Build character device demos
cd /app/cdev_demo/v4l2/
make

Running ISP Tuning

cd /app/tuning_tool/
./run_tuning.sh

GPIO Pin Reference

• Pin 37 - BOARD编码模式下的 GPIO37 (对应 WPI 25)
• Uses Hobot.GPIO library (D-Robotics GPIO wrapper)
• Modes: GPIO.BOARD, GPIO.BCM, GPIO.WPI

Important Notes

1. ⚠️ ALWAYS use system Python - /usr/bin/python3.10, NOT conda
2. Hardware Access: Most operations require root privileges or proper hardware permissions
3. Pin Numbers: Use BOARD encoding (physical pin numbers) or BCM encoding (Broadcom)
4. Library: Uses Hobot.GPIO which is D-Robotics' GPIO wrapper
5. Cleanup: Always call GPIO.cleanup() when done
6. Models: Pre-trained models are in /app/model/basic/
7. Scope: This skill covers /app folder resources only. For system commands and other RDK X5 features, refer to official documentation.

Use Cases

• Control LEDs, motors, relays via GPIO
• Read button/switch states
• Communicate with I2C sensors (temperature, IMU, etc.)
• Interface with SPI devices
• Capture video from camera modules (USB, MIPI)
• Video encoding/decoding
• RTSP video streaming
• Run AI inference (YOLO, SSD, CenterNet, segmentation)
• ISP camera tuning
• Video pipeline processing

---

Note: This skill focuses on /app folder sample resources. It is not a complete RDK X5 operation guide.