Data format: Semantic segmentation

Introduce

Semantic segmentation is an image problem that identifies content and position in an image. Usually, it is necessary to identify all pixels of the target in the image. When defining data formats. Each image corresponds to an image annotation. One-to-one correspondence. The pixels in the image represent the labels for the corresponding pixels in the image.

In the annotated image. The background is marked as0, The annotations for each type are respectively1, 2, 3... and so on.

Example

train_meta.csv

image_Label,image_Source
label/001.jpg,images/001.jpg
label/002.jpg,images/002.jpg

FAQ

1. The size and orientation of the source image and the annotated image should be completely consistent