SimpleDet

Overview

SimpleDet is a custom object-detection toolkit built on top of OpenMMLab components. It is designed for satellite and related computer-vision workloads where you need reproducible workflows for:

• training and fine-tuning detection models
• inference and evaluation
• experiment-friendly project-level packaging

Installation

From source:

git clone https://github.com/sirbastiano/MDet.git
cd MDet
python -m pip install .

For OpenMMLab runtime extras:

python -m pip install ".[openmmlab]"

Install directly from PyPI:

python -m pip install simpledet

Quick Start

Install in editable mode while iterating:

python -m pip install -e .

Sanity check:

python -m simpledet --version

Core entry points

• simpledet.train: training entry points
• simpledet.detect: inference entry points
• simpledet.evaluate: evaluation utilities
• simpledet.cli: command-line checks and diagnostics

Repository structure

• simpledet/: installable Python package (package code and model utilities)
• MyConfigs/: dataset/model configuration sources used in experiments
• studies/, notebooks/: exploratory analyses and research workflows
• runscripts/: executable training/inference launch scripts
• build/: generated build output (not intended for direct edits)

Publishing

Use the project Makefile for deterministic releases.

Recommended one-shot publish

make publish

This runs the following steps:

1. bootstrap — installs/updates build and publish tooling
2. build — builds wheel and sdist
3. check — validates distribution metadata
4. uploads to PyPI via Twine

Test publish / manual controls

make publish-test   # publish to TestPyPI using the same prepare steps
make bootstrap      # install/upgrade build + upload tooling
make build          # build dist artifacts
make check          # validate dist artifacts
make install        # install project from local metadata
make clean          # clean build artifacts

make publish and make publish-test require valid Twine credentials (TWINE_USERNAME and TWINE_PASSWORD, or token-based auth).

Contributing

1. Create a branch.
2. Make your changes.
3. Open a PR with a clear summary.

License

This project is licensed under the MIT License.