DEVAModelSize

Enums · AiDotNet.Enums

Defines the backbone size variants for DEVA (Decoupled Video Segmentation).

For Beginners

DEVA separates task-specific image segmentation from temporal propagation, enabling flexible video segmentation with any image segmentation backbone.

How It Works

Reference: Cheng et al., “Tracking Anything with Decoupled Video Segmentation”, ICCV 2023.

Fields

Field Summary
DEVAModelSize Base Standard variant with ResNet-50 temporal module.
DEVAModelSize Large Large variant with Swin-L temporal module.
int value__