Skip to content

Training model cho X-AnyLabeling

Chuẩn bị dữ liệu

  • Dữ liệu label có format VOC: Ví dụ coco8.zip

Upload datasets

Training model hỗ trợ gắn label

  • Truy cập vào https://hub.ultralytics.com/models
  • Chọn Train Model
  • Chọn dataset đã upload Upload datasets
  • Chọn model đã được ultralytics Training Upload datasets
  • Làm theo hướng dẫn để Traning model (Training trên Google Colab) Upload datasets
  • Sau khi traning xong thì download model ONNX Upload datasets

Config AI model

  • Copy file model ONNX vừa download vào 1 folder va đổi tên thành model.onnx
  • Tạo file config: model.yaml
  • Đay là config mẫu cho model OCR biển số xe:
yaml
type: yolov8
name: plate-number-OCR-r20230520
display_name: Plate Number OCR
model_path: model.onnx # path to model file
nms_threshold: 0.45
confidence_threshold: 0.25
classes:
  - 0
  - 1
  - 2
  - 3
  - 4
  - 5
  - 6
  - 7
  - 8
  - 9
  - A
  - B
  - C
  - D
  - E
  - F
  - G
  - H
  - I
  - J
  - K
  - L
  - M
  - N
  - O
  - P
  - Q
  - R
  - S
  - T
  - U
  - V
  - W
  - X
  - Y
  - Z

Sử dụng AI trong X-AnyLabeling

  • Sau khi đã config AI model, chúng ta có thể sử dụng AI model trong X-AnyLabeling Use AI Model
  • Sau khi load custom model, chúng ta có thể sử dụng AI model để gắn label cho ảnh Use AI Model
    • Có thể chọn gắn label tự động từng ảnh hoặc gắn label tự động cho toàn bộ ảnh
  • Export Annotations theo format mong muốn Use AI Model