-
Notifications
You must be signed in to change notification settings - Fork 0
/
yolo_retrain.py
35 lines (28 loc) · 1.36 KB
/
yolo_retrain.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
# %%
from ultralytics import YOLO
# Load a model
#model = YOLO('yolov8n.yaml') # build a new model from YAML
model = YOLO('yolov8n.pt') # load a pretrained model (recommended for training)
#model = YOLO('yolov8n.yaml').load('yolov8n.pt') # build from YAML and transfer weights
# %%
# Run batched inference on a list of images
results = model(['pipe/train/images/DJI_20240220122723_0022_D.JPG'], device = 0) # return a list of Results objects
# Process results list
for result in results:
boxes = result.boxes # Boxes object for bounding box outputs
masks = result.masks # Masks object for segmentation masks outputs
keypoints = result.keypoints # Keypoints object for pose outputs
probs = result.probs # Probs object for classification outputs
result.show() # display to screen
result.save(filename='result.jpg') # save to disk
# Print the class labels and the bounding box coordinates
labels = boxes.cls # a tensor of shape (N,)
coords = boxes.xyxy # a tensor of shape (N, 4)
print(labels)
print(coords)
# %%
# Train the model
results = model.train(data='pipe.yaml',epochs=100, imgsz=640, device=0, batch = 5) # train a new model for 100 epochs
# %%
#load the model from runs\detect\train6
model = YOLO('runs/detect/train3/weights/best.pt') # load a pretrained model (recommended for training)