README-pytorch-to-coreml.md

Converting models trained using CVNets to CoreML

For conversion, we assume that you are using MAC OS machine.

Classification networks

We can convert the classification models using the following command

export EVAL_DIR='' # Location of results 
export CKPT_NAME='' # Name of the pre-trained model weight file (e.g., checkpoint_ema.pt)
cvnets-convert --common.config-file "${EVAL_DIR}/config.yaml" --common.results-loc $EVAL_DIR --model.classification.pretrained "${EVAL_DIR}/${CKPT_NAME}"  --conversion.coreml-extn mlmodel

Detection networks

We can convert the detection models trained on MS-COCO (81 classes, including background) using the following command

export EVAL_DIR='' # Location of results 
export CKPT_NAME='' # Name of the pre-trained model weight file (e.g., checkpoint_ema.pt)
cvnets-convert --common.config-file "${EVAL_DIR}/config.yaml" --common.results-loc $EVAL_DIR --model.detection.pretrained "${EVAL_DIR}/${CKPT_NAME}"  --conversion.coreml-extn mlmodel --model.detection.n-classes 81

Segmentation networks

We can convert the segmentation models trained on the PASCAL VOC 2012 dataset (21 classes, including background) using the following command

export EVAL_DIR='' # Location of results 
export CKPT_NAME='' # Name of the pre-trained model weight file (e.g., checkpoint_ema.pt)
cvnets-convert --common.config-file "${EVAL_DIR}/config.yaml" --common.results-loc $EVAL_DIR --model.segmentation.pretrained "${EVAL_DIR}/${CKPT_NAME}"  --conversion.coreml-extn mlmodel --model.segmentation.n-classes 21