How to train a custom dataset DDR-NET semantic-segmentation

[Bananaspirit]

💡 Your Question

I found a file in the repository, cityscapes_ddrnet.yaml. Where it is described in detail how to train ddrnet on the cityscapes dataset.

# Instructions:
#   0. Make sure that the data is stored in dataset_params.dataset_dir or add "dataset_params.data_dir=<PATH-TO-DATASET>" at the end of the command below (feel free to check ReadMe)
#   1. Move to the project root (where you will find the ReadMe and src folder)
#   2. Run the command:
#      DDRNet23:        python -m super_gradients.train_from_recipe --config-name=cityscapes_ddrnet
#      DDRNet23-Slim:   python -m super_gradients.train_from_recipe --config-name=cityscapes_ddrnet architecture=ddrnet_23_slim
#      DDRNet39:        python -m super_gradients.train_from_recipe --config-name=cityscapes_ddrnet architecture=ddrnet_39
# Note: add "checkpoint_params.checkpoint_path=<ddrnet23-backbone-pretrained-path>" to use pretrained backbone
#
#  Validation mIoU - Cityscapes, training time:
#      DDRNet23:        input-size: [1024, 2048]     mIoU: 80.26     4 X RTX A5000, 12 H
#      DDRNet23-Slim:   input-size: [1024, 2048]     mIoU: 78.01     4 X RTX A5000, 9 H
#      DDRNet39:        input-size: [1024, 2048]     mIoU: 81.32     4 X RTX A5000, 15 H
#
#  Official git repo:
#      https://github.com/ydhongHIT/DDRNet
#
#  Pretrained checkpoints:
#      Backbones- downloaded from the author's official repo.
#       https://deci-pretrained-models.s3.amazonaws.com/ddrnet/imagenet_pt_backbones/ddrnet23_bb_imagenet.pth
#       https://deci-pretrained-models.s3.amazonaws.com/ddrnet/imagenet_pt_backbones/ddrnet23_slim_bb_imagenet.pth
#       https://deci-pretrained-models.s3.amazonaws.com/ddrnet/imagenet_pt_backbones/ddrnet39_bb_imagenet.pth
#
#      Logs, tensorboards and network checkpoints:
#       DDRNet23:       https://deci-pretrained-models.s3.amazonaws.com/ddrnet/cityscapes/ddrnet23/
#       DDRNet23-Slim:  https://deci-pretrained-models.s3.amazonaws.com/ddrnet/cityscapes/ddrnet23_slim/
#       DDRNet39:       https://deci-pretrained-models.s3.amazonaws.com/ddrnet/cityscapes/ddrnet39/
#
#  Learning rate and batch size parameters, using 4 RTX A5000 with DDP:
#      DDRNet23:        input-size: [1024, 1024]     initial_lr: 0.0075    batch-size: 6 * 4gpus = 24
#      DDRNet23-Slim:   input-size: [1024, 1024]     initial_lr: 0.0075    batch-size: 6 * 4gpus = 24
#      DDRNet39:        input-size: [1024, 1024]     initial_lr: 0.0075    batch-size: 6 * 4gpus = 24
#
#  Comments:
#      * Pretrained backbones were used.

But I don't understand how to train drnet on my own dataset and what format should it be? Cityscapes or COCO

Versions

No response

[Bananaspirit]

On the project page I found the following examples: 1) Segmentation Quick Start 2) Segmentation Transfer Learning 3) How to Connect Custom Dataset

so the question is no longer relevant