Place compiled .tflite files here. They are committed to the repository
so deploy.bat and deploy.sh can use them without needing a Linux machine.
| File | Size | Notes |
|---|---|---|
yolo26n_int8_edgetpu.tflite |
~3–4 MB | Edge TPU compiled (primary) |
yolo26n_int8.tflite |
~3–4 MB | CPU fallback |
The edgetpu_compiler only runs on x86_64 Linux. Use the included Docker
setup to compile from any OS (Windows, macOS, Linux):
# From the yolo-detection-2026-coral-tpu/ root:
bash docker/compile.shOr with Docker Compose:
docker compose -f docker/docker-compose.yml run --rm coral-compilerSee docker/README.md for full instructions.
git add models/*.tflite
git commit -m "feat(coral-tpu): add compiled yolo26n edgetpu model (320x320 INT8)"
git pushIf no EdgeTPU model is present, deploy.bat / deploy.sh will download
ssd_mobilenet_v2_coco_quant_postprocess_edgetpu.tflite as a functional
fallback. This is SSD MobileNet (not YOLO 2026), but confirms the TPU
pipeline works before the YOLO model is compiled.