# RangiLyu/nanodet

NanoDet-Plus⚡Super fast and lightweight anchor-free object detection model. 🔥Only 980 KB(int8) / 1.8MB (fp16) and run 97FPS on cellphone🔥

Repository: https://github.com/RangiLyu/nanodet
Canonical: https://ross.abutalabs.com/products/nanodet
Language: Python
License: Apache-2.0
License Family: permissive
Topics: deep-neural-networks, deep-learning, object-detection, anchor-free, ncnn, shufflenet, pytorch, mnn, repvgg, openvino, efficientnet, android, model-zoo, nanodet-plus, nanodet
Last push: 2024-08-08T02:52:55+00:00

## Health v2 (maintenance only)
Score: 23/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 8, longevity 100
- inputs: {"age_days": 2144, "days_push": 755, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 6252, forks 1118 (observed 2026-08-28T04:09:41.016575+00:00)

## What it is
NanoDet-Plus is a super fast, lightweight anchor-free object detection model implemented in PyTorch, with model sizes as small as 980KB (INT8) and real-time performance on mobile ARM CPUs. It supports easy deployment to multiple inference backends including ncnn, MNN, and OpenVINO, with an Android demo included.

## Use cases
- run real-time object detection on mobile devices
- deploy a tiny object detection model on edge hardware
- train a lightweight anchor-free detector on limited GPU memory
- detect objects in images with a small footprint model
- convert a detection model to ncnn or OpenVINO for deployment
- build an Android app with on-device object detection
- benchmark lightweight object detection models on COCO

## When to choose
- you need real-time object detection on mobile or embedded ARM CPUs
- model size and latency are critical constraints
- you want an easy-to-deploy detector with ncnn/MNN/OpenVINO support
- you have limited GPU memory for training a detection model

## When to avoid
- you need maximum accuracy and can afford large models like YOLO-X or DINO
- you need segmentation, keypoints, or other tasks beyond bounding-box detection
- you only run inference on high-end servers where model size doesn't matter
- you need frequent updates or active community support

## Facets
- artifact type: library
- maturity: stable
- function: machine-learning, deep-learning, computer-vision, image-processing
- domain: deep-learning, computer-vision, image-processing, mobile-development
- platform: python, cpp, cross-platform
- tags: object-detection, anchor-free, lightweight-models, edge-deployment, ncnn, on-device-inference, pytorch, model-zoo, real-time-inference, android, gpu

## Member repositories
- RangiLyu/nanodet (main) score 23

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:41.016575+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:46:36.121925+00:00, confidence not recorded.
  - readme: https://github.com/RangiLyu/nanodet (fetched 2026-08-28T04:09:41.016575+00:00, sha 3a0aa71b2c99)
- Data as of 2026-08-30T08:39:29.467469+00:00.