Skip to content

11-image-model-training-vision.md

Latest commit

 

History

History
16 lines (10 loc) · 828 Bytes

11-image-model-training-vision.md

File metadata and controls

16 lines (10 loc) · 828 Bytes
Raw

Previous | Table of Contents | Next

11. Image Model Training (Vision)

Workflow:

  1. Create project (classification/object detection)
  2. Upload & tag images (min recommended: 5 images/tag; production: 50+ per tag)
  3. Train iterations; evaluate metrics; publish iteration to prediction resource
  4. For object detection include bounding boxes/regions

Best practices: balanced tags, image variety, negative examples.

Previous | Table of Contents | Next