SimpleDet Docs

Installation

Install the package from source, add the CPU runtime extras, then verify the suite and native runtime entry points.

Overview Quickstart Core Concepts API Reference

Use this page first if commands from the other docs are failing and you are not yet sure the environment is valid.

On this page: Requirements Install from source Verify Notes

Requirements

  • Python >=3.10
  • PyTorch runtime compatible with your CPU or CUDA environment
  • simpledet[cpu] for the public package runtime
  • timm if you want suite-driven backbone swapping

Install from source

git clone https://github.com/sirbastiano/SimpleDet.git
cd SimpleDet
python -m pip install -e .
python -m pip install ".[cpu]"

The bundled extras currently include cpu for runtime dependencies, geo for geospatial helpers, and plots for plotting utilities.

Verify the runtime

python -m simpledet --version
python -m simpledet --check-runtime
python - <<'PY'
from simpledet.detectors.data import list_formats
from simpledet.suite import build_detector, compile_native_detector_plan

print(list_formats())
spec = build_detector("retinanet", encoder="resnet18.a1_in1k", num_classes=2, in_channels=3)
print(compile_native_detector_plan(spec).encoder.type)
PY

Expected output shape:

0.1.0
SimpleDet native runtime stack is available.
['coco', 'csv', 'json', 'voc', 'yolo', ...]
TimmEncoder

Notes

  • The documented training, inference, and evaluation flow is native-only.
  • Use simpledet[cpu] before importing simpledet.native or running the CLI.
  • device="auto" in the lightweight inference helper currently resolves to CPU.

If python -m simpledet --check-runtime fails, fix the environment before debugging runtime code.