YOLOv4-tiny Darknet 目标检测训练实战轻量模型配置、训练与推理这篇教程根据我复现 YOLOv4-tiny Darknet 训练流程时整理重点演示 cuDNN 环境、Darknet 编译、YOLO 数据整理、配置生成和测试集推理。本文整理自我的学习和项目复现过程尽量按实操顺序保留 notebook 的关键步骤同时把数据集获取方式调整为适合中文教程发布的写法。本文会重点跑通以下流程配置 Colab CUDA/cuDNN 环境安装并编译 Darknet从数据集后台获取 YOLO Darknet 格式数据动态生成 YOLOv4-tiny 自定义配置训练模型并随机选择测试图推理如果你正在系统学习目标检测、实例分割、OCR、多目标跟踪或视觉大模型建议收藏本文配套 notebook、示例图片和运行环境说明后续会继续整理。如果环境配置卡住可以在评论区说明具体报错。 文章目录YOLOv4-tiny Darknet 目标检测训练实战轻量模型配置、训练与推理⚙️ 配置 CUDA/cuDNN 安装 Darknet 从数据集后台获取 Darknet 数据 整理训练文件 生成训练配置️ 训练 YOLOv4-tiny 推理工具函数 小结 同系列教程汇总⚙️ 配置 CUDA/cuDNNYOLOv4-tiny Darknet 对 CUDA/cuDNN 配置敏感先检查运行环境。# 检查 CUDA 是否已安装以及当前版本。!/usr/local/cuda/bin/nvcc--version# We need to install the correct cuDNN according to this output# 查看当前 GPU 型号!nvidia-smi# This cell ensures you have the correct architecture for your respective GPU# If you command is not found, look through these GPUs, find the respective# GPU and add them to the archTypes dictionary# Tesla V100# ARCH -gencode archcompute_70,code[sm_70,compute_70]# Tesla K80# ARCH -gencode archcompute_37,codesm_37# GeForce RTX 2080 Ti, RTX 2080, RTX 2070, Quadro RTX 8000, Quadro RTX 6000, Quadro RTX 5000, Tesla T4, XNOR Tensor Cores# ARCH -gencode archcompute_75,code[sm_75,compute_75]# Jetson XAVIER# ARCH -gencode archcompute_72,code[sm_72,compute_72]# GTX 1080, GTX 1070, GTX 1060, GTX 1050, GTX 1030, Titan Xp, Tesla P40, Tesla P4# ARCH -gencode archcompute_61,codesm_61# GP100/Tesla P100 - DGX-1# ARCH -gencode archcompute_60,codesm_60# For Jetson TX1, Tegra X1, DRIVE CX, DRIVE PX - uncomment:# ARCH -gencode archcompute_53,code[sm_53,compute_53]# For Jetson Tx2 or Drive-PX2 uncomment:# ARCH -gencode archcompute_62,code[sm_62,compute_62]importos os.environ[GPU_TYPE]str(os.popen(nvidia-smi --query-gpuname --formatcsv,noheader).read())defgetGPUArch(argument):try:argumentargument.strip()# All Colab GPUsarchTypes{Tesla V100-SXM2-16GB:-gencode archcompute_70,code[sm_70,compute_70],Tesla K80:-gencode archcompute_37,codesm_37,Tesla T4:-gencode archcompute_75,code[sm_75,compute_75],Tesla P40:-gencode archcompute_61,codesm_61,Tesla P4:-gencode archcompute_61,codesm_61,Tesla P100-PCIE-16GB:-gencode archcompute_60,codesm_60}returnarchTypes[argument]exceptKeyError:returnGPU must be added to GPU Commandsos.environ[ARCH_VALUE]getGPUArch(os.environ[GPU_TYPE])print(GPU Type: os.environ[GPU_TYPE])print(ARCH Value: os.environ[ARCH_VALUE]) 安装 Darknet克隆并编译 Darknet开启 GPU、cuDNN 和 OpenCV 支持。%cd/content/%rm-rf darknet# 克隆 Darknet 仓库用于 YOLOv4-tiny 训练!git clone https://github.com/AlexeyAB/darknet.git#install environment from the Makefile%cd/content/darknet/# compute_37, sm_37 for Tesla K80# compute_75, sm_75 for Tesla T4# !sed -i s/ARCH -gencode archcompute_60,codesm_60/ARCH -gencode archcompute_75,codesm_75/g Makefile#install environment from the Makefile#note if you are on Colab Pro this works on a P100 GPU#if you are on Colab free, you may need to change the Makefile for the K80 GPU#this goes for any GPU, you need to change the Makefile to inform darknet which GPU you are running on.!sed-is/OPENCV0/OPENCV1/gMakefile !sed-is/GPU0/GPU1/gMakefile !sed-is/CUDNN0/CUDNN1/gMakefile !sed-is/ARCH -gencode archcompute_60,codesm_60/ARCH ${ARCH_VALUE}/gMakefile !make#download the newly released yolov4-tiny weights%cd/content/darknet !wget https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-tiny.weights !wget https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-tiny.conv.29fromtypesimportSimpleNamespace# 从数据集后台下载 YOLO Darknet 格式数据集后修改 DATASET_DIR 指向解压目录。DATASET_DIR/content/dataset# 修改为数据集后台导出的数据集目录datasetSimpleNamespace(locationDATASET_DIR,version1,namecustom-dataset) 从数据集后台获取 Darknet 数据从数据集后台导出 YOLO Darknet 格式数据并接入训练目录。fromtypesimportSimpleNamespace# 从数据集后台下载 YOLO Darknet 检测 格式数据集后修改 DATASET_DIR 指向解压目录。DATASET_DIR/content/dataset# 修改为数据集后台导出的数据集目录datasetSimpleNamespace(locationDATASET_DIR,version1,namecustom-dataset)# 数据集已在上一单元配置如需更换数据请修改 DATASET_DIR。print(dataset.location) 整理训练文件复制图片、标签和类别文件生成 Darknet 训练所需的 obj.data。# 根据类别数量动态生成配置#we build iteratively from base config files. This is the same file shape as cfg/yolo-obj.cfgdeffile_len(fname):withopen(fname)asf:fori,linenumerate(f):passreturni1num_classesfile_len(dataset.location/train/_darknet.labels)max_batchesnum_classes*2000steps1.8*max_batches steps2.9*max_batches steps_strstr(steps1),str(steps2)num_filters(num_classes5)*3print(writing config for a custom YOLOv4 detector detecting number of classes: str(num_classes))#Instructions from the darknet repo#change line max_batches to (classes*2000 but not less than number of training images, and not less than 6000), f.e. max_batches6000 if you train for 3 classes#change line steps to 80% and 90% of max_batches, f.e. steps4800,5400ifos.path.exists(./cfg/custom-yolov4-tiny-detector.cfg):os.remove(./cfg/custom-yolov4-tiny-detector.cfg)#customize iPython writefile so we can write variablesfromIPython.core.magicimportregister_line_cell_magicregister_line_cell_magicdefwritetemplate(line,cell):withopen(line,w)asf:f.write(cell.format(**globals())) 生成训练配置根据类别数动态修改 filters、classes、max_batches 等配置。%%writetemplate./cfg/custom-yolov4-tiny-detector.cfg[net]# Testing#batch1#subdivisions1# Trainingbatch64subdivisions24width416height416channels3momentum0.9decay0.0005angle0saturation1.5exposure1.5hue.1learning_rate0.00261burn_in1000max_batches{max_batches}policysteps steps{steps_str}scales.1,.1[convolutional]batch_normalize1filters32size3stride2pad1activationleaky[convolutional]batch_normalize1filters64size3stride2pad1activationleaky[convolutional]batch_normalize1filters64size3stride1pad1activationleaky[route]layers-1groups2group_id1[convolutional]batch_normalize1filters32size3stride1pad1activationleaky[convolutional]batch_normalize1filters32size3stride1pad1activationleaky[route]layers-1,-2[convolutional]batch_normalize1filters64size1stride1pad1activationleaky[route]layers-6,-1[maxpool]size2stride2[convolutional]batch_normalize1filters128size3stride1pad1activationleaky[route]layers-1groups2group_id1[convolutional]batch_normalize1filters64size3stride1pad1activationleaky[convolutional]batch_normalize1filters64size3stride1pad1activationleaky[route]layers-1,-2[convolutional]batch_normalize1filters128size1stride1pad1activationleaky[route]layers-6,-1[maxpool]size2stride2[convolutional]batch_normalize1filters256size3stride1pad1activationleaky[route]layers-1groups2group_id1[convolutional]batch_normalize1filters128size3stride1pad1activationleaky[convolutional]batch_normalize1filters128size3stride1pad1activationleaky[route]layers-1,-2[convolutional]batch_normalize1filters256size1stride1pad1activationleaky[route]layers-6,-1[maxpool]size2stride2[convolutional]batch_normalize1filters512size3stride1pad1activationleaky##################################[convolutional]batch_normalize1filters256size1stride1pad1activationleaky[convolutional]batch_normalize1filters512size3stride1pad1activationleaky[convolutional]size1stride1pad1filters{num_filters}activationlinear[yolo]mask3,4,5anchors10,14,23,27,37,58,81,82,135,169,344,319classes{num_classes}num6jitter.3scale_x_y1.05cls_normalizer1.0iou_normalizer0.07iou_lossciou ignore_thresh.7truth_thresh1random0nms_kindgreedynms beta_nms0.6[route]layers-4[convolutional]batch_normalize1filters128size1stride1pad1activationleaky[upsample]stride2[route]layers-1,23[convolutional]batch_normalize1filters256size3stride1pad1activationleaky[convolutional]size1stride1pad1filters{num_filters}activationlinear[yolo]mask1,2,3anchors10,14,23,27,37,58,81,82,135,169,344,319classes{num_classes}num6jitter.3scale_x_y1.05cls_normalizer1.0iou_normalizer0.07iou_lossciou ignore_thresh.7truth_thresh1random0nms_kindgreedynms beta_nms0.6# 查看刚写入的配置文件。#you may consider adjusting certain things#like the number of subdivisions 64 runs faster but Colab GPU may not be big enough#if Colab GPU memory is too small, you will need to adjust subdivisions to 16%cat cfg/custom-yolov4-tiny-detector.cfg!./darknet detector train data/obj.data cfg/custom-yolov4-tiny-detector.cfg yolov4-tiny.conv.29-dont_show-map#If you get CUDA out of memory adjust subdivisions above!#adjust max batches down for shorter training above# 定义图片显示工具函数defimShow(path):importcv2importmatplotlib.pyplotasplt%matplotlib inline imagecv2.imread(path)height,widthimage.shape[:2]resized_imagecv2.resize(image,(3*width,3*height),interpolationcv2.INTER_CUBIC)figplt.gcf()fig.set_size_inches(18,10)plt.axis(off)#plt.rcParams[figure.figsize] [10, 5]plt.imshow(cv2.cvtColor(resized_image,cv2.COLOR_BGR2RGB))plt.show()️ 训练 YOLOv4-tiny启动 Darknet 训练。如果显存不足优先调整 subdivisions。# 检查权重是否已保存#backup houses the last weights for our detector#(file yolo-obj_last.weights will be saved to the build\darknet\x64\backup\ for each 100 iterations)#(file yolo-obj_xxxx.weights will be saved to the build\darknet\x64\backup\ for each 1000 iterations)#After training is complete - get result yolo-obj_final.weights from path build\darknet\x64\bac!ls backup#if it is empty you havent trained for long enough yet, you need to train for at least 100 iterations 推理工具函数定义图片显示工具方便查看 Darknet 推理结果。#coco.names is hardcoded somewhere in the detector%cp data/obj.names data/coco.names# test 目录中包含可用于测试的图片test_images[fforfinos.listdir(test)iff.endswith(.jpg)]importrandom img_pathtest/random.choice(test_images);#test out our detector!!./darknet detect cfg/custom-yolov4-tiny-detector.cfg backup/custom-yolov4-tiny-detector_best.weights{img_path}-dont-show imShow(/content/darknet/predictions.jpg) 小结这篇教程完整整理了YOLOv4-tiny Darknet 目标检测训练的核心复现流程。实际操作时建议先确认 GPU、依赖版本、数据集路径和模型权重路径再逐段运行 notebook。后续我会继续按源项目顺序整理同系列中的目标检测、实例分割、OCR、多目标跟踪和视觉大模型教程。 同系列教程汇总Google Gemini 3.5 Flash 零样本目标检测教程从提示词到可视化结果GLM-OCR 文档识别实战教程从验证码、公式到车牌 OCRRF-DETR ByteTrack 多目标跟踪实战教程从命令行到 Python 视频轨迹可视化SAM 3 图像分割实战教程文本、框和点提示的多种分割方式SAM 3 视频分割实战教程用文本提示分割并跟踪视频中的目标YOLOv4-tiny Darknet 目标检测训练实战轻量模型配置、训练与推理-本文