This repository provides codes for "Polarization-sensitive optical coherence tomography with deep learning for detecting the local distribution of osteoarthritis severities".
In regression case, we experiment on two types of regression labels: coarse labels and dense labels.
For coarse label regression:
GPU_ID=0
python main.py --img_type='phase' --epoch=150 --gpu_id=${GPU_ID} --model_select='vgg16' --step_size 20 --gamma 0.2
python main.py --img_type='phase' --epoch=150 --gpu_id=${GPU_ID} --model_select='resnet18' --step_size 20 --gamma 0.2
python main.py --img_type='phase' --epoch=150 --gpu_id=${GPU_ID} --model_select='densenet121' --step_size 20 --gamma 0.2
python main.py --img_type='phase' --epoch=150 --gpu_id=${GPU_ID} --model_select='mobilenetv2' --step_size 20 --gamma 0.2
Similarly, you can experiment on "intensity" images by changing "phase" to "intensity".
For dense label regression:
GPU_ID=0
python main_dense.py --img_type='phase' --epoch=150 --gpu_id=${GPU_ID} --model_select='vgg16' --step_size 20 --gamma 0.2
python main_dense.py --img_type='phase' --epoch=150 --gpu_id=${GPU_ID} --model_select='resnet18' --step_size 20 --gamma 0.2
python main_dense.py --img_type='phase' --epoch=150 --gpu_id=${GPU_ID} --model_select='densenet121' --step_size 20 --gamma 0.2
python main_dense.py --img_type='phase' --epoch=150 --gpu_id=${GPU_ID} --model_select='mobilenetv2' --step_size 20 --gamma 0.2
For PS-OCT classification:
GPU_ID=0
python main_class.py --img_type='phase' --num_class 2 --epoch=150 --gpu_id=${GPU_ID} --model_select='vgg16' --step_size 7 --gamma 0.2
python main_class.py --img_type='phase' --num_class 2 --epoch=150 --gpu_id=${GPU_ID} --model_select='resnet18' --step_size 7 --gamma 0.2
python main_class.py --img_type='phase' --num_class 2 --epoch=150 --gpu_id=${GPU_ID} --model_select='densenet121' --step_size 7 --gamma 0.2
python main_class.py --img_type='phase' --num_class 2 --epoch=150 --gpu_id=${GPU_ID} --model_select='mobilenetv2' --step_size 7 --gamma 0.2