訓練/驗證/預測

Colab train example

https://colab.research.google.com/drive/1cA0i8jSyuPx3CKWhJbIi_Z_b1FEEV5y3?usp=sharing

訓練

# Start training from a pretrained *.pt model!yolo segment train data=coco8-seg.yaml model=yolo26n-seg.pt epochs=100 imgsz=640
# Build a new model from YAML, transfer pretrained weights to it and start training!yolo segment train data=coco8-seg.yaml model=yolo26n-seg.yaml pretrained=yolo26n-seg.pt epochs=100 imgsz=640
# Build a new model from YAML and start training from scratch!yolo segment train data=coco8-seg.yaml model=yolo26n-seg.yaml epochs=100 imgsz=640

驗證

On Colab

!yolo segment val model=path/to/best.pt data=./dataset.yaml
!yolo segment val model=/content/runs/segment/train/weights/best.pt data=/content/dataset.yaml

On Windows

yolo segment val model=C:/Sam/yolo_seg_train/wbc_yolo_seg/runs/segment/train/weights/best.pt data=C:/Sam/yolo_seg_train/wbc_yolo_seg/dataset.yaml imgsz=640
On MAC/Linus# default datasetyolo segment val model=yolo26n-seg.pt # val official model
# custom datasetyolo segment val model=path/to/best.pt data=./dataset.yaml

Test set validation

Ultralytics 8.4.51 🚀 Python-3.12.13 torch-2.10.0+cu128 CUDA:0 (Tesla T4, 14913MiB)YOLO26n-seg summary (fused): 139 layers, 2,690,444 parameters, 0 gradients, 9.0 GFLOPsval: Fast image access ✅ (ping: 0.0±0.0 ms, read: 2580.5±1088.6 MB/s, size: 926.0 KB)val: Scanning /content/wbc_yolo_seg/BASOPHILS_train/test/labels... 35 images, 0 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 35/35 1.1Kit/s 0.0sval: New cache created: /content/wbc_yolo_seg/BASOPHILS_train/test/labels.cache Class Images Instances Box(P R mAP50 mAP50-95) Mask(P R mAP50 mAP50-95): 100% ━━━━━━━━━━━━ 3/3 1.6s/it 4.9s all 35 41 0.676 0.636 0.729 0.681 0.676 0.636 0.729 0.687 3 4 4 1 0.334 0.828 0.729 1 0.334 0.828 0.768 5 8 12 0.547 0.905 0.762 0.742 0.547 0.905 0.762 0.742 6 6 6 0.297 0.667 0.628 0.626 0.297 0.667 0.628 0.593 7 6 6 1 0.275 0.551 0.492 1 0.275 0.551 0.518 8 11 13 0.538 1 0.874 0.815 0.538 1 0.874 0.816Speed: 7.3ms preprocess, 34.7ms inference, 0.0ms loss, 4.2ms postprocess per imageResults saved to /content/runs/segment/val-2💡 Learn more at https://docs.ultralytics.com/modes/val

Download pre-train models

https://docs.ultralytics.com/models/yolo26

!yolo segment predict model=yolo26n-seg.pt source='https://ultralytics.com/images/bus.jpg' # predict with official model!yolo segment predict model=path/to/best.pt source='https://ultralytics.com/images/bus.jpg' # predict with custom model