As per https://developer.qualcomm.com/software/qualcomm-neural-processing-sdk/learning-resources/image-segmentation-deeplab-neural-processing-sdk/deeplab-v3-neural-processing-sdk-ubuntu

the inputs to Deeplab network have to undergo preprocessing to convert them into .RAW format
the code preprocess.py take input images and outputs the .raw images

After that the .raw images need to be grayscaled. the grayscale.py takes care of that

to download the deeplab model

wget http://download.tensorflow.org/models/deeplabv3_mnv2_pascal_train_aug_2018_01_29.tar.gz
tar -xzvf deeplabv3_mnv2_pascal_train_aug_2018_01_29.tar.gz

To convert the .pb file to .dlc format, Activate the virtual environment which was used during snpe installation

snpe-tensorflow-to-dlc –graph deeplabv3_mnv2_pascal_train_aug/frozen_inference_graph.pb -i sub_7 1,513,513,3 --out_node ArgMax --dlc deeplabv3.dlc --allow_unconsumed_nodes

For Quantizing this model using snpe Post Training Quantization (PTQ)

Put raw images a .txt format

find $(pwd)/preprocessed -name '*.raw' > input_list.txt

Quantize

snpe-dlc-quantize --input_dlc model.dlc --input_list input_list.txt --output_dlc quantized_model.dlc