The first integer number (0) is the object class id. For this dataset, the class id 0 refers to the class “using mask” and the class id 1 refers to the “without mask” class. The following float numbers are the xywh bounding box coordinates. As one can see, these coordinates are normalized to [0, 1].
Under the directory /datasets , create a new folder for the MASK dataset. Then, copy the downloaded dataset under this folder. Example: /datasets/dataset_mask/archive/obj/
The dataset is indeed a bunch of images and respective annotation files:
Visualize Train Dataset image with Boundary Box and Label
Under the working space ( YOLOv8/ ) , create the following python file ( visualizeLabel.py) to view images and labels.
## Visualize B.Box and Label on Train Dataset
import cv2
image_path = 'datasets/dataset_mask/archive/obj/2-with-mask'
image = cv2.imread(image_path + '.jpg')
class_list = ['using mask', 'without mask']
colors = [(0, 255, 0), (0, 255, 255)]
height, width, _ = image.shape
T=[]
with open(image_path + '.txt', "r") as file1:
for line in file1.readlines():
split = line.split(" ")
# getting the class id
class_id = int(split[0])
color = colors[class_id]
clazz = class_list[class_id]
# getting the xywh bounding box coordinates
x, y, w, h = float(split[1]), float(split[2]), float(split[3]), float(split[4])
# re-scaling xywh to the image size
box = [int((x - 0.5*w)* width), int((y - 0.5*h) * height), int(w*width), int(h*height)]
cv2.rectangle(image, box, color, 2)
cv2.rectangle(image, (box[0], box[1] - 20), (box[0] + box[2], box[1]), color, -1)
cv2.putText(image, class_list[class_id], (box[0], box[1] - 5), cv2.FONT_HERSHEY_SIMPLEX, .5, (0,0,0))
cv2.imshow("output", image)
cv2.waitKey()
You will see this result
Step 3. Split Dataset
The YOLO training process will use the training subset to actually learn how to detect objects. The validation dataset is used to check the model performance during the training.
We need to split this data into two groups for training model: training and validation.
About 90% of the images will be copied to the folder /training/.
The remaining images (10% of the full data) will be saved in the folder /validation/.
For the inference dataset, you can use any images with people wearing mask.
Under the working directory create the following python file split_data.py.
from ultralytics import YOLO
def train():
# Load a pretrained YOLO model
model = YOLO('yolov8n.pt')
# Train the model using the 'maskdataset.yaml' dataset for 3 epochs
results = model.train(data='maskdataset.yaml', epochs=3)
if __name__ == '__main__':
train()
Finally, in the end, we have the following output:
Now, confirm that you have a yolov8/runs/detect/train/weights/best.pt file:
Depending on the number of runs, it can be under /train#/weights/best.pt, where #:number of train
For my PC, it was train3
Also, check the output of runs/detect/train#/results.png which demonstrates the model performance indicators during the training:
Step 6. Test the model (Inference)
Now we have our model trained with the Labeled Mask dataset, it is time to get some predictions. This can be easily done using an out-of-the-box YOLOv8 script specially designed for this:
Download a test image here and copy the file under the folder of yolov8/datasets/dataset_mask/images/testing
Create the following python file ( Yolov8_test.py) to test model.
from ultralytics import YOLO
import cv2
def test():
# Load a pretrained YOLO model(Change model directory)
model = YOLO('runs/detect/train4/weights/best.pt')
# Inference Source - a single source(Change directory)
src = cv2.imread("datasets/dataset_mask/images/testing/mask-teens.jpg")
# Perform object detection on an image using the model
result = model.predict(source=src, save=True, save_txt=True) # save predictions as labels
# View result
for r in result:
# print the Boxes object containing the detection bounding boxes
print(r.boxes)
# Plot results image
print("result.plot()")
dst = r.plot() # return BGR-order numpy array
cv2.imshow("result plot", dst)
# Plot the original image (NParray)
print("result.orig_img")
cv2.imshow("result orig", r.orig_img)
# Save results to disk
r.save(filename='result.jpg')
cv2.waitKey(0)
if __name__ == '__main__':
test()
Your result image will be saved under runs/detect/predict#/
