Models API

Segmentation Models

UNet

UNet(in_channels=3, classes=2, encoder_name="resnet34")

Standard U-Net with ResNet encoder.

DeepLabV3Plus

DeepLabV3Plus(in_channels=3, classes=2, encoder_name="resnet50")

For multi-scale context aggregation.

Object Detection

FasterRCNN

FasterRCNN(num_classes=10, backbone="resnet50_fpn")

Registry

Models can be instantiated via registry:

from ununennium.models import create_model

model = create_model("unet", encoder_name="efficientnet-b0")