無料GitHub で公開

yolo-detection-2026-coral-tpu-win-wsl

Google Coral Edge TPU — real-time object detection natively via Windows WSL

SharpAI/DeepCamera

インストール方法を見る

含まれるファイル（37）

SKILL.md4.7 KB
_config.yml26 B
.gitignore1.3 KB
.travis.yml1.5 KB
CODE_OF_CONDUCT.md1.9 KB
config.yaml2.2 KB
CONTRIBUTING.md1.6 KB
Contributions.md1.0 KB
deploy.bat3.7 KB
docker_out.log8.7 KB
docker-compose.yml1.3 KB
docker/compile.sh2.7 KB
docker/docker-compose.yml1.4 KB
docker/Dockerfile3.2 KB
docker/README.md2.3 KB
input.json100 B
install_usbipd.bat203 B
install_wsl.bat163 B
libedgetpu.so.11.1 MB
LICENSE1.0 KB
models/README.md1.2 KB
models/ssd_mobilenet_v2_coco_quant_postprocess_edgetpu.tflite6.6 MB
models/ssd_mobilenet_v2_coco_quant_postprocess.tflite5.9 MB
models/yolo26n_full_integer_quant_edgetpu.tflite4.2 MB
models/yolo26n_int8.tflite2.7 MB
output.txt694 B
README.md4.2 KB
requirements.txt701 B
scripts/compile_model_colab.py3.1 KB
scripts/compile_model.py5.2 KB
scripts/detect.py20.9 KB
scripts/detect.py.bak24.4 KB
scripts/install_pycoral.py2.3 KB
scripts/tpu_probe.py3.6 KB
scripts/wsl_wrapper.cjs2.1 KB
skills.json8.4 KB
test2.jpg97.3 KB

SKILL.md（原文）

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Coral TPU Object Detection (Windows WSL)

Real-time object detection natively utilizing the Google Coral Edge TPU accelerator on your local hardware via Windows Subsystem for Linux (WSL). Detects 80 COCO classes (person, car, dog, cat, etc.) with ~4ms inference on 320x320 input.

Requirements

Google Coral USB Accelerator (USB 3.0 port recommended)
WSL2 installed and running on Windows
usbipd-win installed on the Windows host

How It Works

┌─────────────────────────────────────────────────────┐
│ Host (Aegis-AI on Windows)                          │
│   frame.jpg → /tmp/aegis_detection/                 │
│   stdin  ──→ ┌──────────────────────────────┐       │
│              │ WSL Container / Environment   │       │
│              │   detect.py                   │       │
│              │   ├─ loads _edgetpu.tflite     │       │
│              │   ├─ reads frame from disk     │       │
│              │   └─ runs inference on TPU    │       │
│   stdout ←── │   → JSONL detections          │       │
│              └──────────────────────────────┘       │
│   USB ──→ usbipd-win bridge to WSL                  │
└─────────────────────────────────────────────────────┘

Aegis writes camera frame JPEG to shared /tmp/aegis_detection/ workspace
Sends frame event via stdin JSONL to the WSL Python instance
detect.py invokes PyCoral and executes natively on the mapped USB Edge TPU inside Linux
Returns detections event via stdout JSONL back to Windows Host

Performance

Input Size	Inference	On-chip	Notes
320x320	~4ms	100%	Fully on TPU, best for real-time
640x640	~20ms	Partial	Some layers on CPU (model segmented)

Cooling: The USB Accelerator aluminum case acts as a heatsink. If too hot to touch during continuous inference, it will thermal-throttle. Consider active cooling or clock_speed: standard.

Installation

Windows (WSL)

Run deploy.bat — this will:

Verify usbipd is installed and bind the 18d1:9302 and 1a6e:089a Edge TPU hardware IDs.
Setup a Python virtual environment exclusively within WSL.
Install the Edge TPU libraries and dependencies within the WSL boundary.
Auto-attach the device using usbipd seamlessly during invocation.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

ログインしてレビューを書く

同じリポジトリのスキル

概要と使いどころ