This repository extends the Keras example code for Multiclass semantic segmentation using DeepLabV3+. While the original example utilized a ResNet50 backbone, my work focuses on adapting the network to support various backbones available in tensorflow.keras.applications.
Currently, the model can be used with the following backbones:
ResNet50ResNet101ResNet50V2ResNet101V2DenseNet121DenseNet169MobileNetMobileNetV2VGG16VGG19
In my experiments, I found the following backbones were ineffective:
ConvNeXtSmallConvNeXtTinyConvNeXtBaseConvNeXtLargeEfficientNetB0EfficientNetB1EfficientNetB2EfficientNetB3EfficientNetB4EfficientNetB5EfficientNetB6EfficientNetB7EfficientNetV2B0EfficientNetV2B1EfficientNetV2B2EfficientNetV2B3NASNetLarge