SimpleDet Docs

Notebook and visual guide

The repository uses notebooks as the didactic visual layer. Each notebook explains one stage of the workflow, from selecting a backbone to training and evaluating through the pipeline.

Overview Quickstart Core Concepts API Reference

Use this page when you want to navigate the notebooks and visual assets instead of reading the prose docs first.

Package showcase

notebooks/Tools/simpledet_showcase.ipynb is executed with saved outputs and demonstrates the public API without external datasets.

Create and train

notebooks/Tools/create_train_object_detector.ipynb connects backbone, neck, dense head, RPN, and ROI head recipes into runnable detectors, then trains, tests, and runs inference.

Backbone catalog

notebooks/Tools/suite_00_backbone_catalog.ipynb explores encoder discovery and compiled configs.

Dense detectors

notebooks/Tools/suite_01_dense_detector.ipynb shows backbone swapping and dense head adaptation.

ROI detectors

notebooks/Tools/suite_02_roi_detector.ipynb shows Mask R-CNN style assembly and ROI channel patching.

Transformer detectors

notebooks/Tools/suite_03_transformer_detector.ipynb covers ChannelMapper and decoder embed-dimension adaptation.

Training

notebooks/Tools/suite_04_pipeline_training.ipynb shows the canonical training entry point.

Inference and evaluation

notebooks/Tools/suite_05_inference_evaluation.ipynb keeps one detector spec across the whole lifecycle.

Supplemental SVG assets

The repository also keeps simple SVG workflow assets under docs/assets/. These are useful for presentations, but the HTML pages and notebooks are the authoritative docs surface.