# dog-qiuqiu/MobileNet-Yolo

MobileNetV2-YoloV3-Nano: 0.5BFlops 3MB HUAWEI P40: 6ms/img, YoloFace-500k:0.1Bflops 420KB:fire::fire::fire:

Repository: https://github.com/dog-qiuqiu/MobileNet-Yolo
Canonical: https://ross.abutalabs.com/products/mobilenet-yolo
Language: C
License: NOASSERTION
License Family: other
Topics: cnn, yolov3, yolo, mobilenetv2, mobilenet-yolo, ncnn, cv, object-detection, computer-vision, deep-learning, darknet, mnn, face-detection, ncnn-model, mnn-framework, landmark, landmark-detection
Last push: 2021-02-06T02:16:46+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2268, "days_push": 2035, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1744, forks 278 (observed 2026-08-28T04:05:30.576498+00:00)

## What it is
A collection of ultra-lightweight YOLOv3-based object detection models (MobileNetV2-YOLOv3-Lite/Nano, YoloFace) designed for mobile and embedded inference with NCNN and MNN. It includes Darknet training configurations, pre-trained weights, and C samples for detection and landmark tasks.

## Use cases
- run real-time object detection on mobile phones
- deploy a tiny face detection model under 500KB
- train a lightweight YOLOv3 model in Darknet and deploy with NCNN
- detect human pose keypoints on edge devices
- benchmark small detection models on ARM CPUs
- build an embedded camera app with low-FLOPS detection

## When to choose
- you need extremely small and fast object detection models for mobile or embedded ARM devices
- you want NCNN or MNN deployment of YOLO-style detectors
- you need a minimal face detection model with tiny weight size

## When to avoid
- you need the latest state-of-the-art detection models, since the project is no longer updated and points to Yolo-Fastest
- you require GPU training with Darknet group convolutions on Pascal GPUs, which is known to be problematic
- you need a maintained project with active support and recent releases

## Facets
- artifact type: library
- maturity: abandoned
- function: machine-learning, computer-vision, image-processing
- domain: computer-vision, deep-learning, mobile-development
- platform: cross-platform, cpp, python
- tags: object-detection, yolov3, mobilenetv2, face-detection, ncnn, mnn, darknet, lightweight-models, edge-inference, landmark-detection, android, linux

## Member repositories
- dog-qiuqiu/MobileNet-Yolo (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:30.576498+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 2026-08-30T03:29:19.050644+00:00, confidence not recorded.
  - readme: https://github.com/dog-qiuqiu/MobileNet-Yolo (fetched 2026-08-28T04:05:30.576498+00:00, sha 937730b99915)
- Data as of 2026-08-30T08:39:29.467469+00:00.