环境配置

1. 安装conda，官网下载安装即可。
2. 创建环境

conda create -n yolov11 python=3.9 -y
conda activate yolov11

3. 安装PyTorch（MacBook Pro环境，使用MPS模式）

pip install torch torchvision torchaudio

4. 安装Ultralytics库，这是YOLOv11的核心库。

pip install ultralytics

5. 验证环境
   安装完成后，需要确认MPS加速是否可用。启动Python，运行以下代码：

python
import torch
import ultralytics

print(f"PyTorch version: {torch.__version__}")
print(f"MPS (GPU) available: {torch.backends.mps.is_available()}")
print(f"Ultralytics version: {ultralytics.__version__}")

如果看到 MPS (GPU) available: True，说明你的Mac已经可以启用GPU加速了。

训练模型

from ultralytics import YOLO
import os


def train_yolo():
    # 初始化模型（使用预训练权重）
    model = YOLO('yolo11n.pt')  # 可选: yolo11n.pt, yolo11s.pt, yolo11m.pt, yolo11l.pt, yolo11x.pt

    # 训练模型
    results = model.train(
        data='../configs/my_dataset.yaml',  # 数据集配置路径
        epochs=100,  # 训练轮数
        patience=10,  # 早停耐心值
        batch=16,  # 批次大小
        imgsz=640,  # 输入图像尺寸
        device='mps',  # 使用Mac GPU加速
        workers=0,  # Mac上建议设为0
        lr0=0.01,  # 初始学习率
        weight_decay=0.0005,  # 权重衰减
        save=True,  # 保存训练结果
        exist_ok=True,  # 覆盖现有训练结果
        name='my_yolo11_model'  # 训练运行名称
    )

    return results


if __name__ == '__main__':
    train_yolo()

开始训练：

/Users/hogworts/miniconda3/envs/yolo11/bin/python /Users/hogworts/PycharmProjects/my-yolo-proj/scripts/training.py 
Downloading https://github.com/ultralytics/assets/releases/download/v8.3.0/yolo11n.pt to 'yolo11n.pt': 100% ━━━━━━━━━━━━ 5.4MB 10.0MB/s 0.5s
Ultralytics 8.3.217 🚀 Python-3.9.24 torch-2.8.0 MPS (Apple M2)
engine/trainer: agnostic_nms=False, amp=True, augment=False, auto_augment=randaugment, batch=16, bgr=0.0, box=7.5, cache=False, cfg=None, classes=None, close_mosaic=10, cls=0.5, compile=False, conf=None, copy_paste=0.0, copy_paste_mode=flip, cos_lr=False, cutmix=0.0, data=../configs/my_dataset.yaml, degrees=0.0, deterministic=True, device=mps, dfl=1.5, dnn=False, dropout=0.0, dynamic=False, embed=None, epochs=100, erasing=0.4, exist_ok=True, fliplr=0.5, flipud=0.0, format=torchscript, fraction=1.0, freeze=None, half=False, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, imgsz=640, int8=False, iou=0.7, keras=False, kobj=1.0, line_width=None, lr0=0.01, lrf=0.01, mask_ratio=4, max_det=300, mixup=0.0, mode=train, model=yolo11n.pt, momentum=0.937, mosaic=1.0, multi_scale=False, name=my_yolo11_model, nbs=64, nms=False, opset=None, optimize=False, optimizer=auto, overlap_mask=True, patience=10, perspective=0.0, plots=True, pose=12.0, pretrained=True, profile=False, project=None, rect=False, resume=False, retina_masks=False, save=True, save_conf=False, save_crop=False, save_dir=/Users/hogworts/PycharmProjects/my-yolo-proj/scripts/runs/detect/my_yolo11_model, save_frames=False, save_json=False, save_period=-1, save_txt=False, scale=0.5, seed=0, shear=0.0, show=False, show_boxes=True, show_conf=True, show_labels=True, simplify=True, single_cls=False, source=None, split=val, stream_buffer=False, task=detect, time=None, tracker=botsort.yaml, translate=0.1, val=True, verbose=True, vid_stride=1, visualize=False, warmup_bias_lr=0.1, warmup_epochs=3.0, warmup_momentum=0.8, weight_decay=0.0005, workers=0, workspace=None
Matplotlib is building the font cache; this may take a moment.
Overriding model.yaml nc=80 with nc=3

                   from  n    params  module                                       arguments                     
  0                  -1  1       464  ultralytics.nn.modules.conv.Conv             [3, 16, 3, 2]                 
  1                  -1  1      4672  ultralytics.nn.modules.conv.Conv             [16, 32, 3, 2]                
  2                  -1  1      6640  ultralytics.nn.modules.block.C3k2            [32, 64, 1, False, 0.25]      
  3                  -1  1     36992  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2]                
  4                  -1  1     26080  ultralytics.nn.modules.block.C3k2            [64, 128, 1, False, 0.25]     
  5                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]              
  6                  -1  1     87040  ultralytics.nn.modules.block.C3k2            [128, 128, 1, True]           
  7                  -1  1    295424  ultralytics.nn.modules.conv.Conv             [128, 256, 3, 2]              
  8                  -1  1    346112  ultralytics.nn.modules.block.C3k2            [256, 256, 1, True]           
  9                  -1  1    164608  ultralytics.nn.modules.block.SPPF            [256, 256, 5]                 
 10                  -1  1    249728  ultralytics.nn.modules.block.C2PSA           [256, 256, 1]                 
 11                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']          
 12             [-1, 6]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 13                  -1  1    111296  ultralytics.nn.modules.block.C3k2            [384, 128, 1, False]          
 14                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']          
 15             [-1, 4]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 16                  -1  1     32096  ultralytics.nn.modules.block.C3k2            [256, 64, 1, False]           
 17                  -1  1     36992  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2]                
 18            [-1, 13]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 19                  -1  1     86720  ultralytics.nn.modules.block.C3k2            [192, 128, 1, False]          
 20                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]              
 21            [-1, 10]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 22                  -1  1    378880  ultralytics.nn.modules.block.C3k2            [384, 256, 1, True]           
 23        [16, 19, 22]  1    431257  ultralytics.nn.modules.head.Detect           [3, [64, 128, 256]]           
YOLO11n summary: 181 layers, 2,590,425 parameters, 2,590,409 gradients, 6.4 GFLOPs

Transferred 448/499 items from pretrained weights
Freezing layer 'model.23.dfl.conv.weight'
train: Fast image access ✅ (ping: 0.0±0.0 ms, read: 416.0±152.6 MB/s, size: 148.1 KB)
train: Scanning /Users/hogworts/PycharmProjects/my-yolo-proj/datasets/my_dataset/labels/train... 8 images, 0 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 8/8 636.9it/s 0.0s
train: New cache created: /Users/hogworts/PycharmProjects/my-yolo-proj/datasets/my_dataset/labels/train.cache
val: Fast image access ✅ (ping: 0.0±0.0 ms, read: 322.6±177.7 MB/s, size: 75.0 KB)
val: Scanning /Users/hogworts/PycharmProjects/my-yolo-proj/datasets/my_dataset/labels/val... 4 images, 0 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 4/4 4.0Kit/s 0.0s
val: New cache created: /Users/hogworts/PycharmProjects/my-yolo-proj/datasets/my_dataset/labels/val.cache
Plotting labels to /Users/hogworts/PycharmProjects/my-yolo-proj/scripts/runs/detect/my_yolo11_model/labels.jpg... 
optimizer: 'optimizer=auto' found, ignoring 'lr0=0.01' and 'momentum=0.937' and determining best 'optimizer', 'lr0' and 'momentum' automatically... 
optimizer: AdamW(lr=0.001429, momentum=0.9) with parameter groups 81 weight(decay=0.0), 88 weight(decay=0.0005), 87 bias(decay=0.0)
Image sizes 640 train, 640 val
Using 0 dataloader workers
Logging results to /Users/hogworts/PycharmProjects/my-yolo-proj/scripts/runs/detect/my_yolo11_model
Starting training for 100 epochs...

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      1/100       2.3G      1.267      3.303      1.548         23        640: 100% ━━━━━━━━━━━━ 1/1 0.1it/s 19.3s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 0.1it/s 9.1s
                   all          4          4    0.00513          1      0.643      0.554

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      2/100      2.22G      1.365      3.264      1.613         28        640: 100% ━━━━━━━━━━━━ 1/1 0.6it/s 1.6s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 2.3it/s 0.4s
                   all          4          4     0.0055          1      0.631      0.543

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      3/100      2.21G      1.211      3.237      1.427         26        640: 100% ━━━━━━━━━━━━ 1/1 0.9it/s 1.1s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 2.8it/s 0.4s
                   all          4          4    0.00584          1       0.63      0.558

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      4/100      2.21G      1.104      3.221      1.456         26        640: 100% ━━━━━━━━━━━━ 1/1 0.7it/s 1.5s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 2.0it/s 0.5s
                   all          4          4     0.0062          1       0.64      0.568

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      5/100      2.22G       1.22      3.172      1.468         21        640: 100% ━━━━━━━━━━━━ 1/1 1.1it/s 0.9s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 3.1it/s 0.3s
                   all          4          4    0.00677          1      0.652      0.588

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      6/100      2.21G      1.213      3.256      1.574         21        640: 100% ━━━━━━━━━━━━ 1/1 1.3it/s 0.8s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 3.1it/s 0.3s
                   all          4          4    0.00721          1      0.693      0.627

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      7/100      2.21G      1.234      3.462      1.649         18        640: 100% ━━━━━━━━━━━━ 1/1 1.1it/s 0.9s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 3.4it/s 0.3s
                   all          4          4    0.00836          1       0.76       0.69

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      8/100      2.21G     0.8984      2.921      1.335         26        640: 100% ━━━━━━━━━━━━ 1/1 1.4it/s 0.7s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 4.1it/s 0.2s
                   all          4          4     0.0105          1       0.76      0.686

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      9/100      2.21G     0.8167      3.057      1.274         17        640: 100% ━━━━━━━━━━━━ 1/1 1.0it/s 1.0s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 2.7it/s 0.4s
                   all          4          4      0.013          1      0.818      0.735

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     10/100      2.22G     0.7742      2.837      1.084         22        640: 100% ━━━━━━━━━━━━ 1/1 0.9it/s 1.1s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 1.0it/s 1.1s
                   all          4          4     0.0191          1      0.569      0.493

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     11/100      2.21G     0.7602      2.938      1.153         21        640: 100% ━━━━━━━━━━━━ 1/1 0.7it/s 1.4s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 2.1it/s 0.5s
                   all          4          4     0.0219          1      0.569      0.503

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     12/100      2.22G     0.6936      2.788      1.072         18        640: 100% ━━━━━━━━━━━━ 1/1 1.2it/s 0.8s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 3.1it/s 0.3s
                   all          4          4     0.0257          1      0.487      0.444

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     13/100      2.21G     0.7928      2.902      1.186         26        640: 100% ━━━━━━━━━━━━ 1/1 1.2it/s 0.8s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 2.1it/s 0.5s
                   all          4          4     0.0377          1      0.597      0.553

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     14/100      2.21G       0.86      3.125       1.24         21        640: 100% ━━━━━━━━━━━━ 1/1 1.3it/s 0.8s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 2.1it/s 0.5s
                   all          4          4     0.0496          1      0.581      0.548

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     15/100      2.22G     0.8086      2.718      1.277         16        640: 100% ━━━━━━━━━━━━ 1/1 0.7it/s 1.4s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 1.7it/s 0.6s
                   all          4          4     0.0637          1      0.664      0.613

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     16/100      2.21G     0.8007      2.935      1.216         21        640: 100% ━━━━━━━━━━━━ 1/1 1.1it/s 0.9s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 1.3it/s 0.8s
                   all          4          4     0.0694          1      0.622      0.541

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     17/100      2.21G     0.7634      2.737      1.141         28        640: 100% ━━━━━━━━━━━━ 1/1 0.3it/s 3.8s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 1.5it/s 0.7s
                   all          4          4     0.0665          1      0.581       0.48

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     18/100      2.21G     0.7265      2.721      1.259         20        640: 100% ━━━━━━━━━━━━ 1/1 0.9it/s 1.1s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 3.2it/s 0.3s
                   all          4          4     0.0665          1      0.581       0.48

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     19/100      2.27G     0.8433      2.568      1.112         23        640: 100% ━━━━━━━━━━━━ 1/1 1.2it/s 0.8s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 1.4it/s 0.7s
                   all          4          4     0.0846          1      0.539      0.468
EarlyStopping: Training stopped early as no improvement observed in last 10 epochs. Best results observed at epoch 9, best model saved as best.pt.
To update EarlyStopping(patience=10) pass a new patience value, i.e. `patience=300` or use `patience=0` to disable EarlyStopping.

19 epochs completed in 0.022 hours.
Optimizer stripped from /Users/hogworts/PycharmProjects/my-yolo-proj/scripts/runs/detect/my_yolo11_model/weights/last.pt, 5.5MB
Optimizer stripped from /Users/hogworts/PycharmProjects/my-yolo-proj/scripts/runs/detect/my_yolo11_model/weights/best.pt, 5.5MB

Validating /Users/hogworts/PycharmProjects/my-yolo-proj/scripts/runs/detect/my_yolo11_model/weights/best.pt...
Ultralytics 8.3.217 🚀 Python-3.9.24 torch-2.8.0 MPS (Apple M2)
YOLO11n summary (fused): 100 layers, 2,582,737 parameters, 0 gradients, 6.3 GFLOPs
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 0.6it/s 1.6s
                   all          4          4     0.0133          1      0.802      0.722
                   cat          3          3     0.0244          1      0.608      0.548
                  bird          1          1    0.00216          1      0.995      0.895
Speed: 0.4ms preprocess, 283.6ms inference, 0.0ms loss, 51.0ms postprocess per image
Results saved to /Users/hogworts/PycharmProjects/my-yolo-proj/scripts/runs/detect/my_yolo11_model

进程已结束，退出代码为 0

为什么 会出现EarlyStopping警告？

当前情况解读：
训练在 epoch 9 达到最佳性能

随后10个epoch没有进一步改进

系统自动保存了 best.pt（第9个epoch的模型）

检查模型是否已经收敛

相关代码：

from ultralytics import YOLO

def evaluate_training():
    # 加载最佳模型
    model = YOLO('runs/train/my_yolo11_model/weights/best.pt')
    
    # 在验证集上评估模型
    results = model.val(
        data='../configs/my_dataset.yaml',  # 注意路径调整
        device='mps'
    )
    
    print("=== 模型评估结果 ===")
    print(f"mAP50: {results.box.map50:.4f}")
    print(f"mAP50-95: {results.box.map:.4f}")
    print(f"精确度 (Precision): {results.box.p[0]:.4f}")  # 修正这里
    print(f"召回率 (Recall): {results.box.r[0]:.4f}")     # 修正这里
    print(f"F1分数: {results.box.f1[0]:.4f}")
    
    # 各类别详细结果
    print("\n=== 各类别性能 ===")
    for i, class_name in enumerate(['cat', 'bird']):  # 根据你的类别修改
        print(f"{class_name}: mAP50={results.box.ap50[i]:.4f}, mAP50-95={results.box.ap[i]:.4f}")
    
    # 性能评估
    print("\n=== 性能分析 ===")
    if results.box.map50 > 0.8:
        print("✅ 优秀！模型性能非常好！")
        print("💡 建议：训练已经充分，可以直接使用模型")
    elif results.box.map50 > 0.6:
        print("✅ 良好！模型性能不错")
        print("💡 建议：可以继续训练或直接使用")
    elif results.box.map50 > 0.4:
        print("⚠️ 一般！模型有改进空间")
        print("💡 建议：继续训练或调整参数")
    else:
        print("❌ 较差！需要优化")
        print("💡 建议：检查数据质量或重新训练")
    
    return results

if __name__ == '__main__':
    evaluate_training()

结果是好是坏呢？

/Users/hogworts/miniconda3/envs/yolo11/bin/python /Users/hogworts/PycharmProjects/my-yolo-proj/scripts/evaluate_model.py 
Ultralytics 8.3.217 🚀 Python-3.9.24 torch-2.8.0 MPS (Apple M2)
YOLO11n summary (fused): 100 layers, 2,582,737 parameters, 0 gradients, 6.3 GFLOPs
val: Fast image access ✅ (ping: 0.0±0.0 ms, read: 273.1±174.0 MB/s, size: 75.0 KB)
val: Scanning /Users/hogworts/PycharmProjects/my-yolo-proj/datasets/my_dataset/labels/val.cache... 4 images, 0 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 4/4 124.3Kit/s 0.0s
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 1/1 0.9it/s 1.1s
                   all          4          4     0.0133          1      0.802      0.722
                   cat          3          3     0.0244          1      0.608      0.548
                  bird          1          1    0.00216          1      0.995      0.895
Speed: 1.8ms preprocess, 193.0ms inference, 0.0ms loss, 22.4ms postprocess per image
Results saved to /Users/hogworts/PycharmProjects/my-yolo-proj/scripts/runs/detect/val2
=== 模型评估结果 ===
mAP50: 0.8017
mAP50-95: 0.7219
精确度 (Precision): 0.0244
召回率 (Recall): 1.0000
F1分数: 0.0476

=== 各类别性能 ===
cat: mAP50=0.6083, mAP50-95=0.5483
bird: mAP50=0.9950, mAP50-95=0.8955

=== 性能分析 ===
✅ 优秀！模型性能非常好！
💡 建议：训练已经充分，可以直接使用模型

进程已结束，退出代码为 0

性能是不错，但是可用性很差，下面来根据这几个指标得分分析一下。

模型评估

好的模型（理想情况）：

精确度 > 0.7 ✅
召回率 > 0.8 ✅  
F1分数 > 0.75 ✅
mAP50 > 0.8 ✅

你的模型：

精确度 = 0.0244 ❌（严重问题）
召回率 = 1.0000 ✅（优秀）
F1分数 = 0.0476 ❌（严重问题）
mAP50 = 0.8017 ⚠️（表面好看）

1. 精确度 (Precision) = 0.0244 ❌
   通俗理解：模型说"我找到了目标"，这个说法有多可靠？

情况：模型每说100次"我找到了目标"，只有2.44次是真的

比喻：像是一个保安，看到100个人就说"这些都是小偷"，结果只有2个是真的小偷

结论：非常差，模型产生了大量误报

2. 召回率 (Recall) = 1.0000 ✅
   通俗理解：真实的目标中，模型找到了多少？

情况：所有真实目标都被找到了，一个不漏

比喻：保安抓住了所有真正的小偷，但把很多好人也当成了小偷

结论：完美，但代价是很多误检

3. F1分数 = 0.0476 ❌
   通俗理解：精确度和召回率的综合评分

范围：0-1，越接近1越好

情况：0.0476非常低

结论：很差，整体检测质量不好

4. mAP50 = 0.8017 ⚠️
   通俗理解：在宽松标准下的整体检测准确度

情况：0.80看起来不错

但要注意：这个指标可能被高召回率拉高，掩盖了精确度问题