import sys
import torch
import torch.onnx
sys.path.append('./wseg')
from wseg.core.config import cfg, cfg_from_file
from wseg.models import get_model
def add_model(name, model):
models = {}
models[name] = {}
models[name]['model'] = model
return models
def load(paths, models):
for _, data in models.items():
if data['model'] is not None:
data['model'].load_state_dict(torch.load(paths,map_location=torch.device('cpu')))
print("load weight from", paths)
def pth2onnx(cfg_file, pth_file, output_file):
cfg_from_file(cfg_file)
model = get_model(cfg.NET, num_classes=cfg.TEST.NUM_CLASSES)
models = add_model('enc', model)
load(pth_file, models)
for p in model.parameters():
p.requires_grad = False
model.eval()
input_names = ["image"]
output_names = ["cls", "mask"]
dynamic_axes = {'image': {0: '-1'}, 'cls': {0: '-1'}, 'mask': {0: '-1'}}
dummy_input = torch.randn(1, 3, 1024, 1024)
torch.onnx.export(model, dummy_input, output_file, input_names = input_names, dynamic_axes = dynamic_axes, output_names = output_names, verbose=True, opset_version=11)
if __name__ == "__main__":
cfg_path = sys.argv[1]
pth_path = sys.argv[2]
output_path = sys.argv[3]
pth2onnx(cfg_path, pth_path, output_path)