YOLO Datasets And Training Methods

This project mainly completed the following work:

• It introduces how to make a custom dataset for YOLO and how to train a YOLO model by the custom dataset.
• Some modifications have been made to Yolov5, YOLOV6, Yolov7 and Yolov8, which can be adapted to the custom dataset for training.
• Pre-training weights for the object detection model of YOLO are provided.

1. Making Custom Datasets

（1）Capture images

• If you are using Intel RealSense to capture images, you can check this link Images-Acquisition-With-RealSense.

（2）Install and launch labelimg

• Open a terminal and enter the following command to complete the installation of labelimg.

  pip install labelimg -i https://pypi.tuna.tsinghua.edu.cn/simple

• Launch labelimg by typing labelimg in the terminal and the screen will appear as shown below.

  

（3）Configure the annotation tool

• Find View in the menu and click on Auto Save mode to set the annotation process to autosave mode.

  

• Select PascalVOC to set the format of the dataset.

  

（4）Annotate the dataset

• It is recommended that the custom dataset be set up in the following directory structure. The YoloDataSets is the root directory of the dataset, the folder named images contains the images to be annotated, and the folder named Annotations contains the labels generated by the annotation tool.

  YoloDataSets/
   |——————images/
   |        └——————1.jpg
   |        └——————2.jpg  
   |        └——————3.jpg
   |        └——————...
   |——————Annotations/
   |        └——————1.xml
   |        └——————2.xml  
   |        └——————3.xml
   |        └——————...

• Click Open Dir and select the folder named images to find the images to be annotated.

• Click Change Save Dir and select the folder named Annotations to save labels annotated by labelimg。

• Start annotating images, the detailed process is as follows.

  

（5）Convert VOC format to YOLO format

• Save the annotated dataset in the following structure.

  YoloDataSets/
   |——————images/
   |        └——————1.jpg
   |        └——————2.jpg  
   |        └——————3.jpg
   |        └——————...
   |——————Annotations/
   |        └——————1.xml
   |        └——————2.xml  
   |        └——————3.xml
   |        └——————...

• Run DataSet.py by typing the following command in the terminal.

  python DataSet.py --yoloversion yolov5 --trainval_percent 0.9 --train_percent 0.9 --mainpath YoloDataSets --classes ['dog','man']
                                  yolov6                    ···                ···             ····                   ['','',···]
                                  yolov7                    ···                ···             ····                   ['','',···]
                                  yolov8                    ···                ···             ····                   ['','',···]

• The meaning of each parameter in the command is as follows.

  • yoloversion : the version of YOLO, which you can choose YOLOv5, YOLOv6, YOLOv7 and YOLOv8
  • trainval_percent : the total percentage of the training and validation set
  • train_percent : the percentage of training set in training set and validation set
  • mainpath : the root directory of the custom dataset
  • classes : the name of the label, please fill in the list format according to the example

2. Training Model

（1）Training method of YOLOv5

• Enter the following command in the terminal to access the folder named yolov5.

  cd yolov5

• Place the converted dataset in the root directory of yolov5.

• Add the .yaml configuration file named data.yaml to the directory of YoloDataSets with the following content and format.

  path : YoloDataSets
  train: train.txt
  val: val.txt
  test: test.txt
  
  # number of classes
  nc: 2
  
  # class names
  names: ['dog','man']
  

• Run the following command in the terminal, with the parameters adjusted as appropriate.

  python train.py --data YoloDataSets/data.yaml --epochs 300 --weights yolov5n.pt --cfg model/yolov5n.yaml  --batch-size 128
                                                                       yolov5s.pt       model/yolov5s.yaml               64
                                                                       yolov5m.pt       model/yolov5m.yaml               40
                                                                       yolov5l.pt       model/yolov5l.yaml               24
                                                                       yolov5x.pt       model/yolov5x.yaml               16

• The official pre-training weights for object detection are provided as follows.

  Model	size (pixels)	mAPval 50-95	mAPval 50	Speed CPU b1 (ms)	Speed V100 b1 (ms)	Speed V100 b32 (ms)	params (M)	FLOPs @640 (B)
  YOLOv5n	640	28.0	45.7	45	6.3	0.6	1.9	4.5
  YOLOv5s	640	37.4	56.8	98	6.4	0.9	7.2	16.5
  YOLOv5m	640	45.4	64.1	224	8.2	1.7	21.2	49.0
  YOLOv5l	640	49.0	67.3	430	10.1	2.7	46.5	109.1
  YOLOv5x	640	50.7	68.9	766	12.1	4.8	86.7	205.7
  YOLOv5n6	1280	36.0	54.4	153	8.1	2.1	3.2	4.6
  YOLOv5m6	1280	51.3	69.3	887	11.1	6.8	35.7	50.0
  YOLOv5l6	1280	53.7	71.3	1784	15.8	10.5	76.8	111.4
  YOLOv5x6 + [TTA]	1280 1536	55.0 55.8	72.7 72.7	3136 -	26.2 -	19.4 -	140.7 -	209.8 -

（2）Training method of YOLOv6

• Enter the following command in the terminal to access the folder named yolov6.

  cd yolov6

• Place the converted dataset in the root directory of yolov6.

• Add the .yaml configuration file named data.yaml to the directory of YoloDataSets with the following content and format.

  train: YoloDataSets/images/train       # train images
  val: YoloDataSets/images/val           # val images
  test: YoloDataSets/images/test
  is_coco: False
  
  # number of classes
  nc: 2
  
  # class names
  names: ['dog','man']
  

• Run the following command in the terminal, with the parameters adjusted as appropriate.

  python tools/train.py --batch 64 --conf configs/yolov6s6_finetune.py --data YoloDataSets/data.yaml --epochs 300  --device 0         
  

• The official pre-training weights for object detection are provided as follows.

  Model	Size	mAPval 0.5:0.95	SpeedT4 trt fp16 b1 (fps)	SpeedT4 trt fp16 b32 (fps)	Params (M)	FLOPs (G)
  YOLOv6-N	640	37.5	779	1187	4.7	11.4
  YOLOv6-S	640	45.0	339	484	18.5	45.3
  YOLOv6-M	640	50.0	175	226	34.9	85.8
  YOLOv6-L	640	52.8	98	116	59.6	150.7
  YOLOv6-N6	1280	44.9	228	281	10.4	49.8
  YOLOv6-S6	1280	50.3	98	108	41.4	198.0
  YOLOv6-M6	1280	55.2	47	55	79.6	379.5
  YOLOv6-L6	1280	57.2	26	29	140.4	673.4

（3）Training method of YOLOv7

• Enter the following command in the terminal to access the folder named yolov7.

  cd yolov7

• Place the converted dataset in the root directory of yolov7.

• Add the .yaml configuration file named data.yaml to the directory of YoloDataSets with the following content and format.

  train: YoloDataSets/train.txt
  val: YoloDataSets/val.txt
  test: YoloDataSets/test.txt
  
  # number of classes
  nc: 2
  
  # class names
  names: ['dog','man']
  

• Run the following command in the terminal, with the parameters adjusted as appropriate.

  # finetune p5 models
  python train.py --workers 8 --device 0 --batch-size 32 --data YoloDataSets/data.yaml --img 640 640 --cfg cfg/training/yolov7-custom.yaml --weights 'yolov7_training.pt' --name yolov7-custom --hyp data/hyp.scratch.custom.yaml
  
  # finetune p6 models
  python train_aux.py --workers 8 --device 0 --batch-size 16 --data YoloDataSets/data.yaml --img 1280 1280 --cfg cfg/training/yolov7-w6-custom.yaml --weights 'yolov7-w6_training.pt' --name yolov7-w6-custom --hyp data/hyp.scratch.custom.yaml        
  

• The official pre-training weights for object detection are provided as follows.

  Model	Test Size	APtest	AP50test	AP75test	batch 1 fps	batch 32 average time
  YOLOv7	640	51.4%	69.7%	55.9%	161fps	2.8ms
  YOLOv7-X	640	53.1%	71.2%	57.8%	114fps	4.3ms
  YOLOv7-W6	1280	54.9%	72.6%	60.1%	84fps	7.6ms
  YOLOv7-E6	1280	56.0%	73.5%	61.2%	56fps	12.3ms
  YOLOv7-D6	1280	56.6%	74.0%	61.8%	44fps	15.0ms
  YOLOv7-E6E	1280	56.8%	74.4%	62.1%	36fps	18.7ms

（4）Training method of YOLOv8

• Enter the following command in the terminal to access the folder named yolov8.

  cd yolov8

• Place the converted dataset in the root directory of yolov8.

• Add the .yaml configuration file named data.yaml to the directory of YoloDataSets with the following content and format.

  path : ../YoloDataSets
  train: train.txt
  val: val.txt
  test: test.txt
  
  # number of classes
  nc: 2
  
  # class names
  names: ['dog','man']
  

• Run the following command in the terminal, with the parameters adjusted as appropriate.

  yolo task=detect mode=train model=yolov8n.yaml data=YoloDataSets/data.yaml batch=28 epochs=300 imgsz=640 workers=32 device=0
                                    yolov8s.yaml
                                    yolov8m.yaml
                                    yolov8l.yaml
                                    yolov8x.yaml
  

• The official pre-training weights for object detection are provided as follows.

  Model	size (pixels)	mAPval 50-95	Speed CPU ONNX (ms)	Speed A100 TensorRT (ms)	params (M)	FLOPs (B)
  YOLOv8n	640	37.3	80.4	0.99	3.2	8.7
  YOLOv8s	640	44.9	128.4	1.20	11.2	28.6
  YOLOv8m	640	50.2	234.7	1.83	25.9	78.9
  YOLOv8l	640	52.9	375.2	2.39	43.7	165.2
  YOLOv8x	640	53.9	479.1	3.53	68.2	257.8