Toybrick
标题:
pt转rknn的问题
[打印本页]
作者:
zhw
时间:
2020-10-15 10:44
标题:
pt转rknn的问题
out = action_model.predict(pts, frame.shape[:
2
])
这是我用人体关键点检测人体行为的模型，模型输入为关键点pts和图片的shape模型结构是这样的：weight_file =
'./Models/TSSTG/tsstg-model.pth'
graph_args = {
'strategy'
:
'spatial'
}
class_names = [
'Standing'
,
'Walking'
,
'Sitting'
,
'Lying Down'
,
'Stand up'
,
'Sit down'
,
'Fall Down'
]
num_class =
len
(class_names)
model = TwoStreamSpatialTemporalGraph(graph_args, num_class).to(device)
model.load_state_dict(torch.load(weight_file,
map_location
=device))
model.eval()这种模型如何转换？？？？？？
print
(
'--> Loading model'
)
weight_file =
'./Models/TSSTG/tsstg-model.pth'
graph_args = {
'strategy'
:
'spatial'
}
class_names = [
'Standing'
,
'Walking'
,
'Sitting'
,
'Lying Down'
,
'Stand up'
,
'Sit down'
,
'Fall Down'
]
num_class =
len
(class_names)
model = TwoStreamSpatialTemporalGraph(graph_args, num_class).to(device)
model.load_state_dict(torch.load(weight_file,
map_location
=device))
model.eval()
input1 = torch.randn(
18
,
2
)
input2 = torch.randn(
1
,
2
)
trace_model = torch.jit.trace(model, ((input1, input2),))
trace_model.save(
'pose.pt'
)
ret = rknn.load_pytorch(
'pose.pt'
,
input_size_list
=[input1, input2])这样会报错，求大神指点！！！！！！！！！
作者:
zhw
时间:
2020-10-15 14:33
求大佬帮忙
欢迎光临 Toybrick (https://t.rock-chips.com/)
Powered by Discuz! X3.3
