Unofficial implementation of RedNet architecture described in paper 'RedNet: Residual Encoder-Decoder Network for indoor RGB-D Semantic Segmentation'. It's a model for "Semantic Segmentation", but fixed it for "Super Resolution" by modifying the output layer's activation function.
- python 3.7+
- tensorflow 2.0.0+
python train.py \
--num_layers 15 \
--dataset dataset/bsd_images \
--num_epoch 1000 \
--train_batch_size 16 \
--valid_batch_size 1
Apache License 2.0