Home / C/C++ / Machine Learning

Yolo universal target detection model combined with EfficientNet-lite

Yolo universal target detection model combined with EfficientNet-lite, the calculation amount is only 230Mflops(0.23Bflops), and the model size is 1.3MB
Information
Category: C/C++ / Machine Learning
Watchers: 4
Star: 27
Fork: 6
Last update: Sep 7, 2020
Resource links
https://github.com/dog-qiuqiu/Yolo-Fastest

image

⚡ Yolo-Fastest ⚡

  • Simple, fast, compact, easy to transplant
  • A real-time target detection algorithm for all platforms
  • The fastest and smallest known universal target detection algorithm based on yolo
  • The speed is 45% faster than mobilenetv2-yolov3-nano, and the parameter amount is reduced by 56%

Evaluating indicator

Network VOC mAP(0.5) Resolution Run Time(Ncnn 1xCore) Run Time(Ncnn 4xCore) FLOPS Weight size
MobileNetV2-YOLOv3-Nano 65.27 320 11.36ms 5.48ms 0.55BFlops 3.0MB
Yolo-Fastest 60.8 320 6.74ms 4.42ms 0.23BFlops 1.3MB
Yolo-Fastest-XL & 320 &ms &ms 0.70BFlops 3.5MB
  • Test platform Kirin 990
  • Suitable for hardware with extremely tight computing resources
  • This model is recommended to do some simple single object detection suitable for simple application scenarios

Test

image

How to Train

Generate a pre-trained model for the initialization of the model backbone

  ./darknet partial yolo-fastest.cfg yolo-fastest.weights yolo-fastest.conv.109 109

Train

  ./darknet detector train *.data yolo-fastest.cfg yolo-fastest.conv.109

Thanks

0 Open More issues
Closed

How to get the pre-trained weights for yolo-fastest-xl ?

Hello qiuqiu,

Is my understanding below right ?
The commands for getting the pre-trained weights for yolo-fastest-xl and training yolo-fastest-xl are :

./darknet partial yolo-fastest-xl.cfg yolo-fastest-xl.weights yolo-fastest-xl.conv.109 109 ./darknet detector train voc.data yolo-fastest-xl.cfg yolo-fastest-xl.conv.109

Thanks.

opened Sep 9, 2020 by arnoldfychen 9
Closed

请问如果我想训练自己的数据，需要怎么更改.cfg文件？

您好！

如果我想训练自己的数据集，需要如何更改.cfg文件？我看这个项目中的.cfg文件格式和darknet yolov4的还是不太一样，比如我的classes数量和预训练的模型不一样，我需要按什么规则更改呢？

opened Sep 20, 2020 by JimXu1989 6
Closed

请问到哪里找yolo-fastest.conv.109？

感谢您的技术分享！！ 请问预训练weights yolo-fastest.conv.109要到哪里下载呢？ 多谢！

opened Sep 12, 2020 by JimXu1989 4
Closed

预训练模型分类错误的问题

我是用预训练模型在训练了10000批次后，发现模型对区域的检测基本都能检测到，但检测归类的class还是预训练模型里的“car,bus.plan”之类的东西，而不是我自定义的分类。我在各种**.names都定义了我自己的分类，voc.data里也指定了name文件，但还是无效，这和我的训练次数或样本数量有关吗？还是使用预训练模型的问题呢？

opened Mar 26, 2021 by wxtt522 2
Closed

Can’t use GPU for training or testing

I used CUDA 10.2，cudnn 7.6.5，GPU NVIDIA GeForce RTX 2060

I have set up in Makefile GPU=1 CUDNN=1 ifeq ($(GPU), 1) COMMON+= -DGPU -I/usr/local/cuda-10.2/include/ CFLAGS+= -DGPU LDFLAGS+= -L/usr/local/cuda-10.2/lib64 -lcuda -lcudart -lcublas -lcurand

C:\Yolo-Fastest\build\darknet\x64> ./darknet.exe detector test ./cfg/voc.data ./cfg/yolo-fastest.cfg ./cfg/yolo-fastest.weights ./data/person.jpg -i 1 -thresh 0.25 -out_filename ./data/person_output.jpg

CUDA status Error: file: c:\yolo-fastest\src\dark_cuda.c : cuda_set_device() : line: 39 : build time: Nov 24 2020 - 18:36:35 CUDA Error: invalid device ordinal

Can anyone help to fix this?

opened Nov 24, 2020 by chenmingguo 2
Closed

darknet2ncnn output

修改voc类别为五类，训练后将cfg和weights转为ncnn模型，但是输出层有两个Yolov3DetectionOutput，转换的时候已经选了合并yolo层，这里是我漏掉什么操作吗？ 直接转换你的模型是没问题的，如下图输出层已经合并 微信截图_20210512110447 但是我训练出来的模型转换就是两个输出 微信截图_20210512110428

期待解答，谢谢

opened May 11, 2021 by wxkkk 1
Open

i need use tflite to detect

hello master, i use your Yolo-fastest to tflite and i want use it so i need use it NET like this: https://github.com/hunglc007/tensorflow-yolov4-tflite/blob/9f16748aa3f45ff240608da4bd9b1216a29127f5/core/yolov4.py#L146

can you help me to code this ? thank you!

opened Apr 8, 2022 by LiaoZihZrong 0
Open

yolo-fastest.cfg

where could i find this file？ thx

opened Feb 10, 2022 by bolin12 2
Open

yolo-fastest-1.1-xl.bin is VOC or COCO

yolo-fastest-1.1-xl.bin is VOC or COCO THanks!

opened Dec 5, 2021 by xiaoxiongzzz 0
Open

In linux, compile darknet as libdarknet.so. I just write a simple yolo_console_dll.cpp to link .so.

when i type the command as g++ -fPIC -std=c++11 -D_GLIBCXX_USE_CXX11_ABI=0 yolo_console_dll.cpp -L./lib/ -I. -ldarknet $(pkg-config --cflags --libs opencv) -o test. The error is /tmp/ccis7rE8.o：in function‘main’： yolo_console_dll.cpp:(.text+0xb11):‘Detector::Detector(std::string, std::string, int)’undefined reference collect2: error: ld returned 1 exit status

look forward your reply

opened Nov 16, 2021 by xyzhai-pro 0
Open

Please update for Cortex-A72 (ARM v8) 64-bit, so that Pi 4 can be optimized.

null

opened Nov 11, 2021 by spinoza1791 0
Open

ej0 m6

目前嘗試不使用預設的detector 權重開始，訓練自建的資料集。結果會遇到nan，特別是在第二段的v3 loss輸出，會觀察到某一段時間後，loss值直接飆上去，然後就爆掉了。連帶影響到第一段v3 loss變化。

請問各位有遇過類似的狀況嘛？

訓練機器偏小，弱弱的一張1050。跑ubuntu 20.04。使用darknet進行訓練yolo-fastest。自訂的訓練集，有50類，每類約有3000個框框。

opened Sep 2, 2021 by akioolin 0

v.1.1.0(Jul 24, 2021)

1.1(Apr 26, 2021)

v0.1(Sep 15, 2020)

v1.0

Contents

    lukemelas A PyTorch implementation of EfficientNet
    reu2018DL All the trained models used while developing YOLO-LITE
    zsef123 A PyTorch implementation of " EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks."
    dog-qiuqiu Based on Yolo's low-power, ultra-lightweight universal target detection algorithm, the parameter is only 250k, and the speed of the smart phone mobile terminal can reach ~300fps+
    rwightman Pretrained EfficientNet, EfficientNet-Lite, MixNet, MobileNetV3 / V2, MNASNet A1 and B1, FBNet, Single-Path NAS
    hunglc007 YOLOv4, YOLOv3, YOLO-tiny Implemented in Tensorflow 2.0, Android. Convert YOLO v4 .weights tensorflow, tensorrt and tflite
    hollance Tiny YOLO for iOS implemented using CoreML but also using the new MPS graph API.
    r4ghu Almost Real-time Object Detection using Apple's CoreML and YOLO v1 -
    universal-model universal-model-vue - Unified state management for Vue (and also for Angular/React/Svelte)
    EdjeElectronics A tutorial showing how to train, convert, and run TensorFlow Lite object detection models on Android devices, the Raspberry Pi, and more!
    mystic123 Implementation of YOLO v3 object detector in Tensorflow (TF-Slim)
    wuzhihao7788 reproduce the YOLO series of papers in pytorch, including YOLOv4, PP-YOLO, YOLOv5，YOLOv3, etc.
    eric612 A caffe implementation of MobileNet-YOLO detection network
    prakharg24 Implementation for the paper 'YOLO-ReT: Towards High Accuracy Real-time Object Detection on Edge GPUs'
    JeiKeiLim YOLO v3 TensorFlow Lite iOS GPU acceleration. Training from scratch and making a GPU accelerated mobile application.
    qubvel Implementation on EfficientNet model. Keras
    eric612 Compare FasterRCNN,Yolo,SSD model with the same dataset
    mingxingtan EfficientNets snapshot
    amitshekhariitbhu Android TensorFlow Lite Machine Learning Example
    couchbase Lightweight, embedded, syncable NoSQL database engine for Android.
    irrustible To futures-lite as futures-lite is to futures: smaller.
    Share this repo

    Related Repos


    Machine Learning
    114
    OpenVINO™ integration with TensorFlow
    openvinotoolkit English | 简体中文 OpenVINO™ integration with TensorFlow This repository contains the source code of OpenVINO™ integration with TensorFlow, designed for T
     
    Machine Learning
    7.8k
    Vowpal Wabbit is a machine learning system which pushes the frontier of machine learning
    VowpalWabbit Vowpal Wabbit is a machine learning system which pushes the frontier of machine learning with techniques such as online, hashing, allreduce, reductions, learning2search, active, and interactive learning
     
    Machine Learning
    3k
    High performance, easy-to-use, and scalable machine learning (ML) package
    aksnzhy High performance, easy-to-use, and scalable machine learning (ML) package, including linear model (LR), factorization machines (FM), and field-aware factorization machines (FFM) for Python and CLI interface.
     
    Machine Learning
    62
    Sunnet is a light deep learning library
    Tyill Sunnet is a light deep learning library. Linux/Windows License ResNet cpp-example for Win Compare with Tensorflow, inference ResNet50. PC: i5-2400, GF
     
    Machine Learning
    769
    Gesture recognition toolkit
    nickgillian The Gesture Recognition Toolkit (GRT) is a cross-platform, open-source, C++ machine learning library designed for real-time gesture recognition.
     
    Machine Learning
    79
    FANN (Fast Artifical Neural Network) binding in Crystal
    NeuraLegion crystal-fann Crystal bindings for the FANN C lib Installation Add this to your application's shard.yml: dependencies: crystal-fann: github: Neur
     
    Machine Learning
    5
    Serve pytorch / torch models using Drogon
    SABER-labs C++ Torch Server Serve torch models as rest-api using Drogon, example included for resnet18 model for Imagenet. Benchmarks show improvement of ~6-10x
     
    Machine Learning
    187
    FANN (Fast Artificial Neural Network Library) bindings for Node.js
    rlidwka node-fann node-fann is a FANN bindings for Node.js. FANN (Fast Artificial Neural Network Library) is a free open source neural network library, which
     
    Machine Learning
    162.1k
    An Open Source Machine Learning Framework for Everyone
    tensorflow Documentation TensorFlow is an end-to-end open source platform for machine learning. It has a comprehensive, flexible ecosystem of tools, libraries, a
     
    Machine Learning
    6
    This repository is a tensorrt deployment of the onsets and frames model, which is implemen...
    xk-wang Onsets and Frames TensorRT inference This repository is a tensorrt deployment of the onsets and frames model, which is implemented using pytorch (http
     
    Machine Learning
    27
    SigLib Digital Signal Processing and Machine Learning Library
    Numerix-DSP siglib SigLib Digital Signal Processing and Machine Learning Library What is SigLib? SigLib is a highly portable DSP and ML library that supports many
     
    Machine Learning
    10
    a naive NCNN implementation of GFPGAN aims at developing Practical Algorithms for Real-wor...
    FeiGeChuanShu GFPGAN-ncnn a naive ncnn implementation of GFPGAN aims at developing Practical Algorithms for Real-world Face Restoration model support: 1.GFPGANClean
     
    Machine Learning
    4
    use c++ train yolov5
    jedi007 LibTorch-YoLoV5-train-detection use c++ train yolov5 Just for learning!!! Just for learning!!! Just for learning!!! Dependencies Windows10 Qt5.1
     
    Machine Learning
    317
    OpenMMLab Model Deployment Framework
    open-mmlab Introduction English | 简体中文 MMDeploy is an open-source deep learning model deployment toolset. It is a part of the OpenMMLab project. Major features F
     
    Machine Learning
    4
    YoloX for a Jetson Nano 4 using ncnn.
    Qengineering YoloX Jetson Nano YoloX with the ncnn framework. Paper: https://arxiv.org/pdf/2107.08430.pdf Special made for a Jetson Nano, see Q-engineering deep le