YOLOv7_Colab4

YOLO V7 on Google Colaboratory3 †

　新しく別の「じゃんけん(グー・チョキ・パー)」のカスタム・データセットを作成し、Google Colaboratory 上の「YOLO V7」を使用して学習モデルを作成、じゃんけんの「グー・チョキ・パー」をリアルタイム検出する。

YOLO V7 on Google Colaboratory3
- カスタムデータによる学習３「じゃんけんの判定（その２）」
- 更新履歴
参考資料

※ 最終更新:2023/08/13　

↑

カスタムデータによる学習３「じゃんけんの判定（その２）」 †

↑

事前準備 †

カスタムデータによる学習２の環境と手順で進める。
ローカルPC への事前準備
・アップデートファイルをダウンロードする
　update_20230718.zip (784MB) <アップデート・データ> をダウンロード

・「update」フォルダ内をすべてプロジェクトのホームディレクトリに配置する（上書き保存）
　「Windows」の場合 ➡ 「/anacondawin」直下
　「Linux」の場合　 ➡ 「~/」ホームディレクトリ直下

↑

学習データの収集＜ローカルマシン＞ †

カメラアプリの設定
静止画撮影

フォルダ「janken2」に分けてファイル名を付け直す

　janken2                         → 撮影データフォルダ
　│
　├─ goo                        → 「グー」の写真フォルダ
　│　　├─ goo_001.jpg
　│　　├─ goo_002.jpg
　│　　︙　　︙
　│　　└─ goo_200.jpg
　├─ choki                      → 「チョキ」の写真フォルダ
　│　　├─ choki_001.jpg
　│　　├─ choki_002.jpg
　│　　︙　　︙
　│　　└─ choki_200.jpg
　└─ par                        → 「パー」の写真フォルダ
　　　　├─ par_001.jpg
　　　　├─ par_002.jpg
　　　　︙　　︙
　　　　└─ par_200.jpg

↑

ラベリング＜ローカルマシン＞ †

「labelImg」を起動する

(py38a) PS > cd /anaconda_win/work/labelImg/janken2/
(py38a) PS > labelImg par predefined_classes.txt par

左サイドバーの「フォーマット」をYOLOに設定する
「View」メニューバーの「Auto Save mode」「Display Labels」にチェックを入れる

ラベリング（アノテーション）をする

1. 'w' キーを押して枠モードにする
2. マウスで枠を囲う
3. ダイアログボックスでクラス名（「par」「goo」「choki」）を選択
4. 「Save」を押して保存する
連続で素早くラベリング（アノテーション）をする方法
・右上の「Use default label」にチェックを入れ「par」(「goo」「choki」) を入力
・「File」→「Open Dir」で入力画像ディレクトリを選択「par」(「goo」「choki」)
・「File」→「Open Save Dir」で出力ディレクトリを選択「par」(「goo」「choki」)
・左手で 'w' キーを押して枠モードにし、右手マウスで枠を囲う、'd' キーを押して自動Save & Next Image に移動・・・を繰り返す
・「par」が全部終われば「goo」「choki」を同様に繰り返す

・「グー・チョキ・パー」600枚をラベリングする

↑

データセットの作成＜ローカルマシン＞ †

ラベリング（アノテーション）が終わったファイルをデータセットとしてフォルダに振り分ける

　janken2_dataset                 → データセット フォルダ
　│
　├─ train                      → トレーニング用データ フォルダ
　│　　├─ images               → 画像フォルダ
　│　　│　　├─ XXXXXXX.jpg    → 画像ファイル
　│　　︙　　︙
　│　　│
　│　　└─ labels               → ラベル フォルダ
　│　　　　　├─ XXXXXXX.txt    → ラベルファイル
　│　　　　　︙
　│
　├─ varid                      → 学習状況の検証用データフォルダ
　│　　├─ images               → 画像フォルダ
　│　　│　　├─ XXXXXXX.jpg    → 画像ファイル
　︙　　︙　　︙
　│　　│
　│　　└─ labels               → ラベル フォルダ
　│　　　　　├─ XXXXXXX.txt    → ラベルファイル
　︙　　　　　︙
　├─ class.txt                  → クラス分類（参考）
　└─ janken_dataset2.yaml       → データセットの構成ファイル

「train/」：「valid/」は一般的に８：２で振り分ける（１６０枚：４０枚）
ランダムに４０ファイルを抽出するプログラムを修正「file_select2.py」＜43行目＞
```
    s_path = 'janken2/'
    d0_path = 'janken2_dataset/valid/images/'
    d1_path = 'janken2_dataset/valid/labels/'
        :
```

トレーニングの構成を指示する「janken_dataset2.yaml」ファイルを作成する

train: ./data/janken_dataset2/train/images
val: ./data/janken_dataset2/valid/images

nc: 3
names: ['goo', 'choki', 'par']

↑

学習 Training＜クラウド・サービス＞ †

GoogleColab を起動する
ノートブック「物体検出_yolov7.ipynb」を選択する
左サイドバーの「ファイル」を選択することで Googleドライブをマウントする
データ・セット「janken2_dataset」をフォルダごとGoogleドライブの「yolov7/data」にアップする
```
yolov7　
  ┗ data
     ┗ janken2_dataset
```
学習するのに必要な情報を記述したyamlファイル「janken2_dataset.yaml」を「yolov7」の直下にアップする
```
yolov7　
  ┠ data
  ┃  ┗ janken2_dataset
  ┗ janken2_dataset.yaml
```
カレントディレクトリを Googleドライブ「yolov7」へ移動する
```
cd /content/drive/MyDrive/yolov7
```
・結果表示
```
/content/drive/MyDrive/yolov7
```

下記のコマンドで学習を実行する
※ 前回の結果から学習回数(epochs)は 60 とする

!python train.py --workers 8 --batch-size 16 --data janken2_dataset.yaml --cfg cfg/training/yolov7x.yaml --weights 'yolov7x.pt' --name yolov7x_custom2_60 --hyp data/hyp.scratch.p5.yaml --epochs 60 --device 0

▼　- log -　GoogleColab Tesla T4

2023-07-12 19:14:14.488870: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2023-07-12 19:14:15.473879: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
YOLOR 🚀 v0.1-122-g3b41c2c torch 2.0.1+cu118 CUDA:0 (Tesla T4, 15101.8125MB)

Namespace(weights='yolov7x.pt', cfg='cfg/training/yolov7x.yaml', data='janken2_dataset.yaml', hyp='data/hyp.scratch.p5.yaml', epochs=60, batch_size=16, img_size=[640, 640], rect=False, resume=False, nosave=False, notest=False, noautoanchor=False, evolve=False, bucket='', cache_images=False, image_weights=False, device='0', multi_scale=False, single_cls=False, adam=False, sync_bn=False, local_rank=-1, workers=8, project='runs/train', entity=None, name='yolov7x_custom2_60', exist_ok=False, quad=False, linear_lr=False, label_smoothing=0.0, upload_dataset=False, bbox_interval=-1, save_period=-1, artifact_alias='latest', freeze=[0], v5_metric=False, world_size=1, global_rank=-1, save_dir='runs/train/yolov7x_custom2_60', total_batch_size=16)
tensorboard: Start with 'tensorboard --logdir runs/train', view at http://localhost:6006/
hyperparameters: lr0=0.01, lrf=0.1, momentum=0.937, weight_decay=0.0005, warmup_epochs=3.0, warmup_momentum=0.8, warmup_bias_lr=0.1, box=0.05, cls=0.3, cls_pw=1.0, obj=0.7, obj_pw=1.0, iou_t=0.2, anchor_t=4.0, fl_gamma=0.0, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, degrees=0.0, translate=0.2, scale=0.9, shear=0.0, perspective=0.0, flipud=0.0, fliplr=0.5, mosaic=1.0, mixup=0.15, copy_paste=0.0, paste_in=0.15, loss_ota=1
wandb: Install Weights & Biases for YOLOR logging with 'pip install wandb' (recommended)
Overriding model.yaml nc=80 with nc=3

                 from  n    params  module                                  arguments                     
  0                -1  1      1160  models.common.Conv                      [3, 40, 3, 1]                 
  1                -1  1     28960  models.common.Conv                      [40, 80, 3, 2]                
  2                -1  1     57760  models.common.Conv                      [80, 80, 3, 1]                
  3                -1  1    115520  models.common.Conv                      [80, 160, 3, 2]               
  4                -1  1     10368  models.common.Conv                      [160, 64, 1, 1]               
  5                -2  1     10368  models.common.Conv                      [160, 64, 1, 1]               
  6                -1  1     36992  models.common.Conv                      [64, 64, 3, 1]                
  7                -1  1     36992  models.common.Conv                      [64, 64, 3, 1]                
  8                -1  1     36992  models.common.Conv                      [64, 64, 3, 1]                
  9                -1  1     36992  models.common.Conv                      [64, 64, 3, 1]                
 10                -1  1     36992  models.common.Conv                      [64, 64, 3, 1]                
 11                -1  1     36992  models.common.Conv                      [64, 64, 3, 1]                
 12[-1, -3, -5, -7, -8]  1         0  models.common.Concat                    [1]                           
 13                -1  1    103040  models.common.Conv                      [320, 320, 1, 1]              
 14                -1  1         0  models.common.MP                        []                            
 15                -1  1     51520  models.common.Conv                      [320, 160, 1, 1]              
 16                -3  1     51520  models.common.Conv                      [320, 160, 1, 1]              
 17                -1  1    230720  models.common.Conv                      [160, 160, 3, 2]              
 18          [-1, -3]  1         0  models.common.Concat                    [1]                           
 19                -1  1     41216  models.common.Conv                      [320, 128, 1, 1]              
 20                -2  1     41216  models.common.Conv                      [320, 128, 1, 1]              
 21                -1  1    147712  models.common.Conv                      [128, 128, 3, 1]              
 22                -1  1    147712  models.common.Conv                      [128, 128, 3, 1]              
 23                -1  1    147712  models.common.Conv                      [128, 128, 3, 1]              
 24                -1  1    147712  models.common.Conv                      [128, 128, 3, 1]              
 25                -1  1    147712  models.common.Conv                      [128, 128, 3, 1]              
 26                -1  1    147712  models.common.Conv                      [128, 128, 3, 1]              
 27[-1, -3, -5, -7, -8]  1         0  models.common.Concat                    [1]                           
 28                -1  1    410880  models.common.Conv                      [640, 640, 1, 1]              
 29                -1  1         0  models.common.MP                        []                            
 30                -1  1    205440  models.common.Conv                      [640, 320, 1, 1]              
 31                -3  1    205440  models.common.Conv                      [640, 320, 1, 1]              
 32                -1  1    922240  models.common.Conv                      [320, 320, 3, 2]              
 33          [-1, -3]  1         0  models.common.Concat                    [1]                           
 34                -1  1    164352  models.common.Conv                      [640, 256, 1, 1]              
 35                -2  1    164352  models.common.Conv                      [640, 256, 1, 1]              
 36                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 37                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 38                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 39                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 40                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 41                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 42[-1, -3, -5, -7, -8]  1         0  models.common.Concat                    [1]                           
 43                -1  1   1640960  models.common.Conv                      [1280, 1280, 1, 1]            
 44                -1  1         0  models.common.MP                        []                            
 45                -1  1    820480  models.common.Conv                      [1280, 640, 1, 1]             
 46                -3  1    820480  models.common.Conv                      [1280, 640, 1, 1]             
 47                -1  1   3687680  models.common.Conv                      [640, 640, 3, 2]              
 48          [-1, -3]  1         0  models.common.Concat                    [1]                           
 49                -1  1    328192  models.common.Conv                      [1280, 256, 1, 1]             
 50                -2  1    328192  models.common.Conv                      [1280, 256, 1, 1]             
 51                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 52                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 53                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 54                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 55                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 56                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 57[-1, -3, -5, -7, -8]  1         0  models.common.Concat                    [1]                           
 58                -1  1   1640960  models.common.Conv                      [1280, 1280, 1, 1]            
 59                -1  1  11887360  models.common.SPPCSPC                   [1280, 640, 1]                
 60                -1  1    205440  models.common.Conv                      [640, 320, 1, 1]              
 61                -1  1         0  torch.nn.modules.upsampling.Upsample    [None, 2, 'nearest']          
 62                43  1    410240  models.common.Conv                      [1280, 320, 1, 1]             
 63          [-1, -2]  1         0  models.common.Concat                    [1]                           
 64                -1  1    164352  models.common.Conv                      [640, 256, 1, 1]              
 65                -2  1    164352  models.common.Conv                      [640, 256, 1, 1]              
 66                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 67                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 68                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 69                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 70                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 71                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 72[-1, -3, -5, -7, -8]  1         0  models.common.Concat                    [1]                           
 73                -1  1    410240  models.common.Conv                      [1280, 320, 1, 1]             
 74                -1  1     51520  models.common.Conv                      [320, 160, 1, 1]              
 75                -1  1         0  torch.nn.modules.upsampling.Upsample    [None, 2, 'nearest']          
 76                28  1    102720  models.common.Conv                      [640, 160, 1, 1]              
 77          [-1, -2]  1         0  models.common.Concat                    [1]                           
 78                -1  1     41216  models.common.Conv                      [320, 128, 1, 1]              
 79                -2  1     41216  models.common.Conv                      [320, 128, 1, 1]              
 80                -1  1    147712  models.common.Conv                      [128, 128, 3, 1]              
 81                -1  1    147712  models.common.Conv                      [128, 128, 3, 1]              
 82                -1  1    147712  models.common.Conv                      [128, 128, 3, 1]              
 83                -1  1    147712  models.common.Conv                      [128, 128, 3, 1]              
 84                -1  1    147712  models.common.Conv                      [128, 128, 3, 1]              
 85                -1  1    147712  models.common.Conv                      [128, 128, 3, 1]              
 86[-1, -3, -5, -7, -8]  1         0  models.common.Concat                    [1]                           
 87                -1  1    102720  models.common.Conv                      [640, 160, 1, 1]              
 88                -1  1         0  models.common.MP                        []                            
 89                -1  1     25920  models.common.Conv                      [160, 160, 1, 1]              
 90                -3  1     25920  models.common.Conv                      [160, 160, 1, 1]              
 91                -1  1    230720  models.common.Conv                      [160, 160, 3, 2]              
 92      [-1, -3, 73]  1         0  models.common.Concat                    [1]                           
 93                -1  1    164352  models.common.Conv                      [640, 256, 1, 1]              
 94                -2  1    164352  models.common.Conv                      [640, 256, 1, 1]              
 95                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 96                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 97                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 98                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
 99                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
100                -1  1    590336  models.common.Conv                      [256, 256, 3, 1]              
101[-1, -3, -5, -7, -8]  1         0  models.common.Concat                    [1]                           
102                -1  1    410240  models.common.Conv                      [1280, 320, 1, 1]             
103                -1  1         0  models.common.MP                        []                            
104                -1  1    103040  models.common.Conv                      [320, 320, 1, 1]              
105                -3  1    103040  models.common.Conv                      [320, 320, 1, 1]              
106                -1  1    922240  models.common.Conv                      [320, 320, 3, 2]              
107      [-1, -3, 59]  1         0  models.common.Concat                    [1]                           
108                -1  1    656384  models.common.Conv                      [1280, 512, 1, 1]             
109                -2  1    656384  models.common.Conv                      [1280, 512, 1, 1]             
110                -1  1   2360320  models.common.Conv                      [512, 512, 3, 1]              
111                -1  1   2360320  models.common.Conv                      [512, 512, 3, 1]              
112                -1  1   2360320  models.common.Conv                      [512, 512, 3, 1]              
113                -1  1   2360320  models.common.Conv                      [512, 512, 3, 1]              
114                -1  1   2360320  models.common.Conv                      [512, 512, 3, 1]              
115                -1  1   2360320  models.common.Conv                      [512, 512, 3, 1]              
116[-1, -3, -5, -7, -8]  1         0  models.common.Concat                    [1]                           
117                -1  1   1639680  models.common.Conv                      [2560, 640, 1, 1]             
118                87  1    461440  models.common.Conv                      [160, 320, 3, 1]              
119               102  1   1844480  models.common.Conv                      [320, 640, 3, 1]              
120               117  1   7375360  models.common.Conv                      [640, 1280, 3, 1]             
121   [118, 119, 120]  1     56144  models.yolo.IDetect                     [3, [[12, 16, 19, 36, 40, 28], [36, 75, 76, 55, 72, 146], [142, 110, 192, 243, 459, 401]], [320, 640, 1280]]
Model Summary: 467 layers, 70828568 parameters, 70828568 gradients

Transferred 630/644 items from yolov7x.pt
Scaled weight_decay = 0.0005
Optimizer groups: 108 .bias, 108 conv.weight, 111 other
train: Scanning 'data/janken2_dataset/train/labels' images and labels... 480 found, 0 missing, 0 empty, 0 corrupted: 100% 480/480 [02:01<00:00,  3.94it/s]
train: New cache created: data/janken2_dataset/train/labels.cache
val: Scanning 'data/janken2_dataset/valid/labels' images and labels... 120 found, 0 missing, 0 empty, 0 corrupted: 100% 120/120 [00:30<00:00,  3.89it/s]
val: New cache created: data/janken2_dataset/valid/labels.cache

autoanchor: Analyzing anchors... anchors/target = 4.19, Best Possible Recall (BPR) = 1.0000
Image sizes 640 train, 640 test
Using 2 dataloader workers
Logging results to runs/train/yolov7x_custom2_60
Starting training for 60 epochs...

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
      0/59     14.4G    0.0681   0.01932   0.02179    0.1092        35       640: 100% 30/30 [01:02<00:00,  2.09s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95:   0% 0/4 [00:00<?, ?it/s]/usr/local/lib/python3.10/dist-packages/torch/functional.py:504: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at ../aten/src/ATen/native/TensorShape.cpp:3483.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:15<00:00,  3.76s/it]
                 all         120         120      0.0217      0.0449      0.0101     0.00163

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
      1/59     14.3G   0.05685   0.01645   0.02203   0.09533        44       640: 100% 30/30 [00:44<00:00,  1.49s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.27it/s]
                 all         120         120      0.0202       0.183      0.0143     0.00231

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
      2/59     14.3G   0.05164   0.01496   0.02144   0.08804        42       640: 100% 30/30 [00:43<00:00,  1.44s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.40it/s]
                 all         120         120       0.111       0.183      0.0777      0.0221

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
      3/59     14.3G   0.05053    0.0141   0.02079   0.08541        66       640: 100% 30/30 [00:43<00:00,  1.44s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.51it/s]
                 all         120         120       0.234       0.673       0.266       0.108

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
      4/59     14.3G   0.05113   0.01154   0.02029   0.08296        48       640: 100% 30/30 [00:42<00:00,  1.43s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.55it/s]
                 all         120         120       0.361       0.558       0.365       0.181

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
      5/59     14.3G   0.05104   0.01062   0.01954   0.08119        36       640: 100% 30/30 [00:43<00:00,  1.44s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.49it/s]
                 all         120         120       0.151       0.217       0.111      0.0353

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
      6/59     14.3G   0.04784   0.01013   0.01921   0.07719        34       640: 100% 30/30 [00:42<00:00,  1.40s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.57it/s]
                 all         120         120       0.362       0.748       0.422       0.153

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
      7/59     14.3G   0.04462  0.008979    0.0176    0.0712        50       640: 100% 30/30 [00:42<00:00,  1.40s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.54it/s]
                 all         120         120       0.369       0.583       0.365       0.174

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
      8/59     14.3G   0.04962  0.009079   0.01714   0.07584        52       640: 100% 30/30 [00:41<00:00,  1.37s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.60it/s]
                 all         120         120       0.411       0.715       0.495        0.25

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
      9/59     14.3G   0.04233  0.009555   0.01745   0.06933        42       640: 100% 30/30 [00:41<00:00,  1.39s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.26it/s]
                 all         120         120       0.511       0.764       0.591       0.314

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     10/59     14.3G   0.04662  0.009044    0.0175   0.07317        47       640: 100% 30/30 [00:42<00:00,  1.41s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.60it/s]
                 all         120         120       0.477       0.642       0.549       0.326

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     11/59     14.3G   0.04523  0.008884   0.01668   0.07079        57       640: 100% 30/30 [00:42<00:00,  1.43s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.28it/s]
                 all         120         120        0.55        0.67        0.61       0.373

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     12/59     14.3G   0.04498  0.008311   0.01595   0.06924        47       640: 100% 30/30 [00:42<00:00,  1.40s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.59it/s]
                 all         120         120       0.418       0.792       0.564       0.271

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     13/59     14.3G   0.04339  0.007621   0.01556   0.06657        37       640: 100% 30/30 [00:40<00:00,  1.34s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.27it/s]
                 all         120         120        0.39       0.567        0.43       0.183

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     14/59     14.3G   0.04663  0.008304   0.01703   0.07196        46       640: 100% 30/30 [00:40<00:00,  1.36s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.42it/s]
                 all         120         120       0.248        0.35       0.187       0.075

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     15/59     14.3G   0.04339  0.008267   0.01685    0.0685        48       640: 100% 30/30 [00:40<00:00,  1.37s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.60it/s]
                 all         120         120        0.54       0.516       0.338       0.212

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     16/59     14.3G     0.037   0.00794   0.01478   0.05972        54       640: 100% 30/30 [00:41<00:00,  1.37s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.54it/s]
                 all         120         120       0.516       0.798       0.608       0.352

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     17/59     14.3G   0.04314   0.00783   0.01439   0.06535        54       640: 100% 30/30 [00:41<00:00,  1.38s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.26it/s]
                 all         120         120       0.377       0.695       0.466       0.266

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     18/59     14.2G   0.03568  0.008158   0.01409   0.05793        39       640: 100% 30/30 [00:41<00:00,  1.38s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.31it/s]
                 all         120         120       0.124       0.241       0.074      0.0186

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     19/59     14.2G    0.0403  0.008222   0.01486   0.06338        39       640: 100% 30/30 [00:42<00:00,  1.41s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.58it/s]
                 all         120         120       0.686     0.00833     0.00123    0.000245

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     20/59     14.2G   0.04804  0.007868   0.01696   0.07287        39       640: 100% 30/30 [00:42<00:00,  1.43s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.51it/s]
                 all         120         120      0.0289       0.367      0.0129      0.0053

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     21/59     14.2G   0.04198  0.008768   0.01731   0.06806        53       640: 100% 30/30 [00:41<00:00,  1.40s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.46it/s]
                 all         120         120    0.000968      0.0417    0.000202     3.4e-05

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     22/59     14.2G   0.03543  0.007735   0.01534   0.05851        37       640: 100% 30/30 [00:40<00:00,  1.34s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.29it/s]
                 all         120         120       0.421       0.462        0.34       0.169

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     23/59     14.2G   0.03544  0.007921   0.01435   0.05771        48       640: 100% 30/30 [00:41<00:00,  1.39s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.45it/s]
                 all         120         120       0.517       0.912       0.682       0.433

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     24/59     14.2G   0.03697  0.007951   0.01438    0.0593        54       640: 100% 30/30 [00:42<00:00,  1.41s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.58it/s]
                 all         120         120        0.49       0.557       0.516       0.303

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     25/59     14.1G   0.03675  0.007778   0.01517    0.0597        42       640: 100% 30/30 [00:42<00:00,  1.40s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.27it/s]
                 all         120         120       0.334       0.324       0.252       0.109

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     26/59     14.1G   0.03648  0.007837   0.01329   0.05761        43       640: 100% 30/30 [00:41<00:00,  1.38s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.58it/s]
                 all         120         120        0.28       0.283       0.207      0.0985

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     27/59     14.1G   0.03634  0.007276   0.01417   0.05778        49       640: 100% 30/30 [00:41<00:00,  1.39s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.60it/s]
                 all         120         120       0.174        0.05      0.0284     0.00969

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     28/59     14.1G   0.04075  0.007394   0.01422   0.06236        53       640: 100% 30/30 [00:40<00:00,  1.36s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.62it/s]
                 all         120         120     0.00932     0.00833    0.000543    0.000167

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     29/59     14.1G   0.03572  0.007363    0.0137   0.05678        54       640: 100% 30/30 [00:41<00:00,  1.38s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.33it/s]
                 all         120         120      0.0479       0.075      0.0112     0.00435

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     30/59     14.1G   0.03766  0.006934   0.01316   0.05775        23       640: 100% 30/30 [00:41<00:00,  1.37s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.52it/s]
                 all         120         120      0.0931       0.122      0.0466      0.0187

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     31/59     14.2G   0.03331  0.007154   0.01285   0.05331        53       640: 100% 30/30 [00:41<00:00,  1.38s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.59it/s]
                 all         120         120       0.639       0.467       0.489       0.312

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     32/59     14.2G   0.03142  0.006452   0.01162   0.04949        50       640: 100% 30/30 [00:40<00:00,  1.36s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.59it/s]
                 all         120         120       0.701       0.854       0.825       0.544

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     33/59     14.2G   0.03462  0.006683   0.01137   0.05267        48       640: 100% 30/30 [00:41<00:00,  1.38s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.29it/s]
                 all         120         120       0.604       0.724       0.704       0.458

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     34/59     14.2G    0.0352  0.006353    0.0116   0.05315        37       640: 100% 30/30 [00:40<00:00,  1.36s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.27it/s]
                 all         120         120       0.829        0.84       0.942       0.672

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     35/59     14.2G   0.03385  0.006672   0.01176   0.05228        44       640: 100% 30/30 [00:43<00:00,  1.45s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.39it/s]
                 all         120         120       0.825       0.838       0.911        0.59

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     36/59     14.2G   0.03679  0.006203   0.01175   0.05474        46       640: 100% 30/30 [00:40<00:00,  1.36s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.22it/s]
                 all         120         120       0.892       0.891       0.963        0.68

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     37/59     14.2G   0.03231  0.006649   0.01071   0.04967        34       640: 100% 30/30 [00:43<00:00,  1.44s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.60it/s]
                 all         120         120       0.933       0.906       0.975       0.708

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     38/59     14.2G   0.03375  0.006575   0.01119   0.05152        39       640: 100% 30/30 [00:41<00:00,  1.39s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.23it/s]
                 all         120         120       0.858       0.833       0.915       0.604

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     39/59     14.2G   0.03002  0.006665   0.01091    0.0476        51       640: 100% 30/30 [00:41<00:00,  1.38s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.51it/s]
                 all         120         120       0.962       0.873       0.976       0.697

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     40/59     14.2G   0.03477  0.006127   0.01094   0.05183        39       640: 100% 30/30 [00:42<00:00,  1.41s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.59it/s]
                 all         120         120       0.892       0.889       0.954       0.672

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     41/59     14.2G    0.0344  0.006646   0.01126    0.0523        47       640: 100% 30/30 [00:41<00:00,  1.37s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.59it/s]
                 all         120         120       0.954       0.967       0.984       0.719

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     42/59     14.2G   0.03319  0.006344   0.01118   0.05071        38       640: 100% 30/30 [00:43<00:00,  1.44s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.58it/s]
                 all         120         120       0.841       0.741       0.805       0.517

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     43/59     14.2G   0.02878  0.006224   0.01048   0.04548        52       640: 100% 30/30 [00:40<00:00,  1.36s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.61it/s]
                 all         120         120       0.968       0.946       0.983       0.693

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     44/59     14.2G   0.03183  0.006496   0.01069   0.04902        54       640: 100% 30/30 [00:41<00:00,  1.38s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.58it/s]
                 all         120         120        0.94       0.891       0.976       0.688

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     45/59     14.2G   0.03003  0.006326   0.01033   0.04669        49       640: 100% 30/30 [00:40<00:00,  1.35s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.42it/s]
                 all         120         120       0.922       0.945       0.973       0.716

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     46/59     14.2G    0.0294  0.006092  0.009538   0.04503        34       640: 100% 30/30 [00:40<00:00,  1.34s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.30it/s]
                 all         120         120       0.943       0.966       0.982       0.739

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     47/59     14.2G   0.03086  0.005871  0.009502   0.04624        37       640: 100% 30/30 [00:42<00:00,  1.43s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.61it/s]
                 all         120         120       0.964       0.966       0.989       0.757

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     48/59     14.2G   0.03091  0.006323   0.01035   0.04758        51       640: 100% 30/30 [00:41<00:00,  1.39s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.29it/s]
                 all         120         120       0.911        0.85       0.933       0.693

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     49/59     14.2G   0.03011  0.006137   0.01022   0.04647        40       640: 100% 30/30 [00:40<00:00,  1.34s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.37it/s]
                 all         120         120       0.967       0.958       0.991       0.756

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     50/59     14.2G   0.02953  0.005935  0.009395   0.04486        49       640: 100% 30/30 [00:43<00:00,  1.46s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.32it/s]
                 all         120         120        0.95       0.974       0.986       0.735

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     51/59     14.2G   0.03174  0.006027   0.01057   0.04834        58       640: 100% 30/30 [00:40<00:00,  1.35s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.56it/s]
                 all         120         120       0.986       0.903       0.985       0.723

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     52/59     14.2G   0.02378   0.00587  0.008013   0.03767        38       640: 100% 30/30 [00:40<00:00,  1.36s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.60it/s]
                 all         120         120        0.92       0.965       0.986       0.761

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     53/59     14.2G   0.02978  0.005794  0.009871   0.04545        43       640: 100% 30/30 [00:42<00:00,  1.41s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.58it/s]
                 all         120         120       0.982       0.958       0.991       0.752

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     54/59     14.2G   0.02775  0.006075  0.009443   0.04327        47       640: 100% 30/30 [00:40<00:00,  1.36s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.48it/s]
                 all         120         120       0.981       0.982       0.994        0.76

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     55/59     14.2G   0.02921  0.005971  0.009573   0.04476        38       640: 100% 30/30 [00:41<00:00,  1.37s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:05<00:00,  1.29s/it]
                 all         120         120       0.987       0.975       0.994       0.747

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     56/59     14.2G   0.02385  0.005905  0.008025   0.03778        39       640: 100% 30/30 [00:42<00:00,  1.40s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.49it/s]
                 all         120         120       0.988        0.99       0.993        0.76

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     57/59     14.2G   0.02787   0.00592  0.008957   0.04275        41       640: 100% 30/30 [00:41<00:00,  1.40s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.60it/s]
                 all         120         120       0.987        0.98       0.994       0.768

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     58/59     14.2G   0.02616  0.005833   0.00942   0.04142        46       640: 100% 30/30 [00:42<00:00,  1.42s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:02<00:00,  1.34it/s]
                 all         120         120       0.982       0.975       0.993       0.758

     Epoch   gpu_mem       box       obj       cls     total    labels  img_size
     59/59     14.2G   0.02654   0.00555  0.009116    0.0412        45       640: 100% 30/30 [00:42<00:00,  1.41s/it]
               Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 4/4 [00:03<00:00,  1.06it/s]
                 all         120         120       0.977       0.983       0.992       0.753
                 goo         120          40       0.991       0.975       0.993       0.694
               choki         120          40       0.952           1        0.99       0.778
                 par         120          40       0.988       0.975       0.994       0.785
60 epochs completed in 0.843 hours.

Optimizer stripped from runs/train/yolov7x_custom2_60/weights/last.pt, 142.1MB
Optimizer stripped from runs/train/yolov7x_custom2_60/weights/best.pt, 142.1MB

・評価指標の推移

Epoch	Precision	Recall	mAP@.5	mAP@.5:.95
0	0.0217	0.0449	0.0101	0.00163
1	0.0202	0.183	0.0143	0.00231
2	0.111	0.183	0.0777	0.0221
3	0.234	0.673	0.266	0.108
4	0.361	0.558	0.365	0.181
6	0.362	0.748	0.422	0.153
7	0.369	0.583	0.365	0.174
8	0.411	0.715	0.495	0.25
9	0.511	0.764	0.591	0.314
10	0.477	0.642	0.549	0.326
11	0.55	0.67	0.61	0.373
12	0.418	0.792	0.564	0.271
13	0.39	0.567	0.43	0.183
14	0.248	0.35	0.187	0.075
15	0.54	0.516	0.338	0.212
16	0.516	0.798	0.608	0.352
17	0.377	0.695	0.466	0.266
18	0.124	0.241	0.074	0.0186
19	0.686	0.00833	0.00123	0.000245
20	0.0289	0.367	0.0129	0.0053
21	0.000968	0.0417	0.000202	0.000034
22	0.421	0.462	0.34	0.169
23	0.517	0.912	0.682	0.433
24	0.49	0.557	0.516	0.303
25	0.334	0.324	0.252	0.109
26	0.28	0.283	0.207	0.0985
27	0.174	0.05	0.0284	0.00969
28	0.00932	0.00833	0.000543	0.000167
29	0.0479	0.075	0.0112	0.00435
30	0.0931	0.122	0.0466	0.0187
31	0.639	0.467	0.489	0.312
32	0.701	0.854	0.825	0.544
33	0.604	0.724	0.704	0.458
34	0.829	0.84	0.942	0.672
35	0.825	0.838	0.911	0.59
36	0.892	0.891	0.963	0.68
37	0.933	0.906	0.975	0.708
38	0.858	0.833	0.915	0.604
39	0.962	0.873	0.976	0.697
40	0.892	0.889	0.954	0.672
41	0.954	0.967	0.984	0.719
42	0.841	0.741	0.805	0.517
43	0.968	0.946	0.983	0.693
44	0.94	0.891	0.976	0.688
45	0.922	0.945	0.973	0.716
46	0.943	0.966	0.982	0.739
47	0.964	0.966	0.989	0.757
48	0.911	0.85	0.933	0.693
49	0.967	0.958	0.991	0.756
50	0.95	0.974	0.986	0.735
51	0.986	0.903	0.985	0.723
52	0.92	0.965	0.986	0.761
53	0.982	0.958	0.991	0.752
54	0.981	0.982	0.994	0.76
55	0.987	0.975	0.994	0.747
56	0.988	0.99	0.993	0.76
57	0.987	0.98	0.994	0.768
58	0.982	0.975	0.993	0.758
59	0.977	0.983	0.992	0.753

・学習結果の評価指標

Class	Precision	Recall	mAP@.5	mAP@.5:.95
all	0.977	0.983	0.992	0.753
goo	0.991	0.975	0.993	0.694
choki	0.952	1	0.99	0.778
par	0.988	0.975	0.994	0.785

※ 用語（参考）→ 詳しくは「参考資料」

名称		意味
P適合率	Precision	正例と予測したもののうち、どれだけ正解だったか
R再現率	Recall	正解が正例のもののうち、どれだけ正例と予測できたか
AP平均適合率	Average Precision	m個の正解ラベルのうち、どのくらいラベルを検出できているかを平均的に表したもの
mAP	mean Average Precision	物体の情報が与えられた時点までの適合率（Precision）の平均であるAP（Average Precision）の平均

およそ50分で正常に終了すると、学習結果は「yolov7/runs/train/yolov7x_custom2_60/」に保存される
・学習結果モデルは「yolov7x_custom2_60/weights」評価指標は「yolov7x_custom2_60」

F1 curve P curve PR curve R curve
学習結果「runs/train/yolov7x_custom2_60/」をローカルマシンにダウンロードする

↑

学習済みモデルを ONNX 形式にコンバートする＜クラウド・サービス＞ †

onnx パッケージをインストールする

!pip install onnx

▼　- log -　GoogleColab Tesla T4

Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Collecting onnx
  Downloading onnx-1.14.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (14.6 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 14.6/14.6 MB 78.2 MB/s eta 0:00:00
Requirement already satisfied: numpy in /usr/local/lib/python3.10/dist-packages (from onnx) (1.22.4)
Requirement already satisfied: protobuf>=3.20.2 in /usr/local/lib/python3.10/dist-packages (from onnx) (3.20.3)
Requirement already satisfied: typing-extensions>=3.6.2.1 in /usr/local/lib/python3.10/dist-packages (from onnx) (4.7.1)
Installing collected packages: onnx
Successfully installed onnx-1.14.0

学習結果モデル「runs/train/yolov7x_custom2_60/weights/best.pt」を onnx 形式に変換する

!python export.py --weights runs/train/yolov7x_custom2_60/weights/best.pt

▼　- log -　GoogleColab Tesla T4

Import onnx_graphsurgeon failure: No module named 'onnx_graphsurgeon'
Namespace(weights='runs/train/yolov7x_custom2_60/weights/best.pt', img_size=[640, 640], batch_size=1, dynamic=False, dynamic_batch=False, grid=False, end2end=False, max_wh=None, topk_all=100, iou_thres=0.45, conf_thres=0.25, device='cpu', simplify=False, include_nms=False, fp16=False, int8=False)
YOLOR 🚀 v0.1-122-g3b41c2c torch 2.0.1+cu118 CPU

Fusing layers... 
IDetect.fuse
Model Summary: 362 layers, 70795920 parameters, 0 gradients

Starting TorchScript export with torch 2.0.1+cu118...
TorchScript export success, saved as runs/train/yolov7x_custom2_60/weights/best.torchscript.pt
CoreML export failure: No module named 'coremltools'

Starting TorchScript-Lite export with torch 2.0.1+cu118...
TorchScript-Lite export success, saved as runs/train/yolov7x_custom2_60/weights/best.torchscript.ptl

Starting ONNX export with onnx 1.14.0...
/content/drive/MyDrive/yolov7/models/yolo.py:582: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  if augment:
/content/drive/MyDrive/yolov7/models/yolo.py:614: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  if profile:
/content/drive/MyDrive/yolov7/models/yolo.py:629: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  if profile:
============= Diagnostic Run torch.onnx.export version 2.0.1+cu118 =============
verbose: False, log level: Level.ERROR
======================= 0 NONE 0 NOTE 0 WARNING 0 ERROR ========================

ONNX export success, saved as runs/train/yolov7x_custom2_60/weights/best.onnx

Export complete (65.71s). Visualize with https://github.com/lutzroeder/netron.

変換結果「runs/train/yolov7x_custom2_60/weights/best.onnx」をローカルマシンにダウンロードする

↑

学習結果で推論を実行する画像編＜ローカルマシン＞ †

学習結果モデル「runs/train/yolov7x_custom2_60/weights/best.pt」を使用する
・テスト画像は「/Images/janken2.jpg」を使用する
・ローカルマシン上の「yolov7/」フォルダの場所は「/work/yolov7-main/」
・推論実行コマンドは「detect.py」を修正した「detect2.py」を使用する

検出コマンドを実行する

(py38a) PS > cd /anaconda_win/work/yolov7-main
(py38a) PS > python detect2.py --weights runs/train/yolov7x_custom2_60/weights/best.pt --conf 0.25 --img-size 640 --source ../../Images/janken2.jpg --view-img

▼　- log -　NVIDIA GeForce GTX 1050 Ti

 Namespace(agnostic_nms=False, augment=False, classes=None, conf_thres=0.25, device='', exist_ok=False, img_size=640, iou_thres=0.45, name='exp', no_trace=False, nosave=False, project='runs/detect', save_conf=False, save_txt=False, source='../../Images/janken2.jpg', update=False, view_img=True, weights=['runs/train/yolov7x_custom2_60/weights/best.pt'])
YOLOR  2023-4-13 torch 2.0.1+cu117 CUDA:0 (NVIDIA GeForce GTX 1050 Ti, 4095.75MB)

Fusing layers...
IDetect.fuse
Model Summary: 362 layers, 70795920 parameters, 0 gradients
 Convert model to Traced-model...
 traced_script_module saved!
 model is traced!

C:\Users\XXXXX\anaconda3\envs\py38a\lib\site-packages\torch\functional.py:504: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\builder\windows\pytorch\aten\src\ATen\native\TensorShape.cpp:3484.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]

3 goos, 3 chokis, 3 pars, Done. (130.7ms) Inference, (6.0ms) NMS, (1.70) fps
 The image with the result is saved in: runs\detect\exp22\janken2.jpg
Results saved to runs\detect\exp22

FPS average:       1.70
Done. (0.642s)

・推論結果の保存先は「yolov7/runs/detect/exp*」（ * は自動的に振られる番号）

↑

学習結果で推論を実行する動画編＜ローカルマシン＞ †

学習結果モデル「runs/train/yolov7x_custom2_60/weights/best.pt」を使用する
・テスト画像は Webカメラを使用する
・ローカルマシン上の「yolov7/」フォルダの場所は「/work/yolov7-main/」
・推論実行コマンドは「detect.py」を修正した「detect2.py」を使用する

検出コマンドを実行する（終了は 'Esc'キー）

(py38a) PS > cd /anaconda_win/work/yolov7-main
(py38a) PS > python detect2.py --weights runs/train/yolov7x_custom2_60/weights/best.pt --conf 0.25 --img-size 640 --source 0

▼　- log -　NVIDIA GeForce GTX 1050 Ti

Namespace(agnostic_nms=False, augment=False, classes=None, conf_thres=0.25, device='', exist_ok=False, img_size=640, iou_thres=0.45, name='exp', no_trace=False, nosave=False, project='runs/detect', save_conf=False, save_txt=False, source='0', update=False, view_img=False, weights=['runs/train/yolov7x_custom2_60/weights/best.pt'])
YOLOR  2023-4-13 torch 2.0.1+cu117 CUDA:0 (NVIDIA GeForce GTX 1050 Ti, 4095.75MB)

Fusing layers...
IDetect.fuse
Model Summary: 362 layers, 70795920 parameters, 0 gradients
 Convert model to Traced-model...
 traced_script_module saved!
 model is traced!

1/1: 0...  success (640x480 at 30.00 FPS).

C:\Users\XXXXX\anaconda3\envs\py38a\lib\site-packages\torch\functional.py:504: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\builder\windows\pytorch\aten\src\ATen\native\TensorShape.cpp:3484.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]

0: Done. (129.6ms) Inference, (0.0ms) NMS, (6.30) fps
0: Done. (130.7ms) Inference, (1.0ms) NMS, (6.20) fps
0: Done. (133.7ms) Inference, (0.0ms) NMS, (7.00) fps
0: 1 goo, Done. (131.6ms) Inference, (7.0ms) NMS, (5.90) fps
0: 1 goo, Done. (131.6ms) Inference, (2.0ms) NMS, (5.90) fps
0: 1 choki, Done. (128.7ms) Inference, (2.0ms) NMS, (6.50) fps
0: 1 choki, Done. (128.6ms) Inference, (2.0ms) NMS, (6.30) fps
0: 1 par, Done. (131.6ms) Inference, (2.0ms) NMS, (6.90) fps
0: 1 par, Done. (130.7ms) Inference, (2.0ms) NMS, (6.30) fps
        :
        :
0: 1 par, Done. (132.6ms) Inference, (2.0ms) NMS, (7.00) fps
0: 1 par, Done. (131.6ms) Inference, (2.0ms) NMS, (6.30) fps
Results saved to runs\detect\exp25

FPS average:       6.40
Done. (58.056s)

・推論結果の保存先は「yolov7/runs/detect/exp*」（ * は自動的に振られる番号）

↑

OpenVINO™ で実行する＜ローカルマシン＞ †

　○ Intel® 製の GPU マシンで動作させるために OpenVINO™ を利用する
　○ 学習済みモデルは「best.pt」を onnx にコンバートした「best.onnx」を使用
　○ 学習済みモデルのパスは「yolov7-main/runs/train/yolov7x_custom2_60/weights/」

カメラ画像入力（GPU） Intel® HD Graphics 530

(py38a) PS > cd /anaconda_win/work/yolov7
(py38a) PS > python object_detect_yolo7.py -m ../yolov7-main/runs/train/yolov7x_custom2_60/weights/best.onnx -l janken.names_jp -i cam -d GPU
Starting..
 - Program title  : Object detection YOLO V7
 - OpenCV version : 4.5.5
 - OpenVINO engine: 2022.1.0-7019-cdb9bec7210-releases/2022/1
 - Input image    : cam
 - Model          : ../yolov7-main/runs/train/yolov7x_custom2_60/weights/best.onnx
 - Device         : GPU
 - Confidence thr : 0.1
 - IOU threshold  : 0.6
 - Label          : janken.names_jp
 - Log level      : 3
 - Title flag     : y
 - Speed flag     : y
 - Processed out  : non
 - Preprocessing  : False
 - Batch size     : 1
 - number of inf  : 1
 - With grid      : False
[ WARN:0@42.597] global D:\a\opencv-python\opencv-python\opencv\modules\videoio\src\cap_msmf.cpp (539) `anonymous-namespace'::SourceReaderCB::~SourceReaderCB terminating async callback
FPS average:       1.40
Finished.

カメラ画像入力（CPU） Intel® Core™ i7-6700

(py38a) PS > cd /anaconda_win/work/yolov7
(py38a) PS > python object_detect_yolo7.py -m ../yolov7-main/runs/train/yolov7x_custom2_60/weights/best.onnx -l janken.names_jp -i cam
Starting..
 - Program title  : Object detection YOLO V7
 - OpenCV version : 4.5.5
 - OpenVINO engine: 2022.1.0-7019-cdb9bec7210-releases/2022/1
 - Input image    : cam
 - Model          : ../yolov7-main/runs/train/yolov7x_custom2_60/weights/best.onnx
 - Device         : CPU
 - Confidence thr : 0.1
 - IOU threshold  : 0.6
 - Label          : janken.names_jp
 - Log level      : 3
 - Title flag     : y
 - Speed flag     : y
 - Processed out  : non
 - Preprocessing  : False
 - Batch size     : 1
 - number of inf  : 1
 - With grid      : False
[ WARN:0@56.366] global D:\a\opencv-python\opencv-python\opencv\modules\videoio\src\cap_msmf.cpp (539) `anonymous-namespace'::SourceReaderCB::~SourceReaderCB terminating async callback
FPS average:       0.90
Finished.

静止画入力（GPU） Intel® HD Graphics 530

(py38a) PS > cd /anaconda_win/work/yolov7
(py38a) PS > > python object_detect_yolo7.py -m ../yolov7-main/runs/train/yolov7x_custom2_60/weights/best.onnx -l janken.names_jp -i ../../Images/janken2.jpg -d GPU
Starting..
 - Program title  : Object detection YOLO V7
 - OpenCV version : 4.5.5
 - OpenVINO engine: 2022.1.0-7019-cdb9bec7210-releases/2022/1
 - Input image    : ../../Images/janken2.jpg
 - Model          : ../yolov7-main/runs/train/yolov7x_custom2_60/weights/best.onnx
 - Device         : GPU
 - Confidence thr : 0.1
 - IOU threshold  : 0.6
 - Label          : janken.names_jp
 - Log level      : 3
 - Title flag     : y
 - Speed flag     : y
 - Processed out  : non
 - Preprocessing  : False
 - Batch size     : 1
 - number of inf  : 1
 - With grid      : False
FPS average:       1.40
Finished.

静止画入力（CPU） Intel® Core™ i7-6700

(py38a) PS > cd /anaconda_win/work/yolov7
(py38a) PS > python object_detect_yolo7.py -m ../yolov7-main/runs/train/yolov7x_custom2_60/weights/best.onnx -l janken.names_jp -i ../../Images/janken2.jpg
Starting..
 - Program title  : Object detection YOLO V7
 - OpenCV version : 4.5.5
 - OpenVINO engine: 2022.1.0-7019-cdb9bec7210-releases/2022/1
 - Input image    : ../../Images/janken2.jpg
 - Model          : ../yolov7-main/runs/train/yolov7x_custom2_60/weights/best.onnx
 - Device         : CPU
 - Confidence thr : 0.1
 - IOU threshold  : 0.6
 - Label          : janken.names_jp
 - Log level      : 3
 - Title flag     : y
 - Speed flag     : y
 - Processed out  : non
 - Preprocessing  : False
 - Batch size     : 1
 - number of inf  : 1
 - With grid      : False
FPS average:       0.90
Finished.

マシンによる処理速度(FPS) の違い

OS	GPU		CPU		実行マシン
Windows10	Intel® HD Graphics 530	1.40	Intel® Core™ i7-6700	0.90	HP EliteDesk800 G2 SFF (第6世代 Core™ i7 CPU搭載デスクトップ)
Windows11	Intel® Iris® Xe Graphics	4.40	Intel® Core™ i7-1260P	2.10	DELL XPS Plus 9320 (第12世代 Core™ i7 CPU搭載ノート)
Ubuntu20.04LTS	Intel® Iris® Xe Graphics	3.70	Intel® Core™ i7-1185G7	1.40	DELL Latitude 7520 (第11世代 Core™ i7 CPU搭載ノート)
Ubuntu20.04LTS	Intel® UHD Graphics	1.50	Intel® Core™ i5-10210U	1.20	Intel® NUC BXNUC10I5FNH (第10世代 Core™ i5 CPU搭載ミニPC)