Модель сегментации Pytorch (.ppt) не конвертируется в CoreML ⇐ Python

[Гость]

Согласно ссылке на статью Apple, нам нужно обернуть модель, чтобы разрешить отслеживание, что выполняется так же.

Код: Выделить всё

class WrappedDeeplabv3Resnet1011(nn.Module):

def __init__(self):
super(WrappedDeeplabv3Resnet1011, self).__init__()
self.model = torch.load('/content/aircraft_best_model.pt',map_location ='cpu').eval()

def forward(self, x):
res = self.model(x)
# extract the tensor we want from the output dictionary
x = res['out']
return x

I see the my model doesn't have the key "Out" но я вижу другие клавиши, как показано ниже

Код: Выделить всё

[{'boxes': tensor([[ 510.2429,  229.1375, 1011.1587,  399.5730],
[ 550.1007,  202.8524, 1047.5089,  376.9215],
[ 457.9409,  196.4182,  947.7454,  412.4210],
[ 333.6804,  204.8605, 1073.0546,  442.6238]],
grad_fn=), 'labels': tensor([1, 2, 3, 1]), 'scores': tensor([0.0870, 0.0631, 0.0587, 0.0531], grad_fn=)}]

when I apply any of these keys as output it throws the error as shown below

Код: Выделить всё

TypeError: list indices must be integers or slices, not str

my model eval() shown like below

Код: Выделить всё

FasterRCNN(
(transform): GeneralizedRCNNTransform(
Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
Resize(min_size=(800,), max_size=1333, mode='bilinear')
)
(backbone): BackboneWithFPN(
(body): IntermediateLayerGetter(
(conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
(bn1): FrozenBatchNorm2d(64, eps=0.0)
(relu): ReLU(inplace=True)
(maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
(layer1): Sequential(
(0): Bottleneck(
(conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): FrozenBatchNorm2d(64, eps=0.0)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): FrozenBatchNorm2d(64, eps=0.0)
(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): FrozenBatchNorm2d(256, eps=0.0)
(relu): ReLU(inplace=True)
(downsample): Sequential(
(0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): FrozenBatchNorm2d(256, eps=0.0)
)
)
(1): Bottleneck(
(conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): FrozenBatchNorm2d(64, eps=0.0)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): FrozenBatchNorm2d(64, eps=0.0)
(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): FrozenBatchNorm2d(256, eps=0.0)
(relu): ReLU(inplace=True)
)
(2): Bottleneck(
(conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): FrozenBatchNorm2d(64, eps=0.0)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): FrozenBatchNorm2d(64, eps=0.0)
(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): FrozenBatchNorm2d(256, eps=0.0)
(relu): ReLU(inplace=True)
)
)
(layer2): Sequential(
(0): Bottleneck(
(conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): FrozenBatchNorm2d(128, eps=0.0)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn2): FrozenBatchNorm2d(128, eps=0.0)
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): FrozenBatchNorm2d(512, eps=0.0)
(relu): ReLU(inplace=True)
(downsample): Sequential(
(0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
(1): FrozenBatchNorm2d(512, eps=0.0)
)
)
(1): Bottleneck(
(conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): FrozenBatchNorm2d(128, eps=0.0)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): FrozenBatchNorm2d(128, eps=0.0)
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): FrozenBatchNorm2d(512, eps=0.0)
(relu): ReLU(inplace=True)
)
(2): Bottleneck(
(conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): FrozenBatchNorm2d(128, eps=0.0)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): FrozenBatchNorm2d(128, eps=0.0)
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): FrozenBatchNorm2d(512, eps=0.0)
(relu): ReLU(inplace=True)
)
(3): Bottleneck(
(conv1):  Conv2d(512, 128, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn1): FrozenBatchNorm2d(128, eps=0.0)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), шаг=(1, 1), заполнение=(1, 1), смещение=False)
(bn2): FrozenBatchNorm2d(128, eps=0.0)
(conv3): Conv2d(128, 512, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn3): FrozenBatchNorm2d(512, eps=0.0)
(relu): ReLU(inplace=True)
)
)
(layer3): Последовательный (
(0): Узкое место (
(conv1) ): Conv2d(512, 256, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn1): FrozenBatchNorm2d(256, eps=0.0)
(conv2 ): Conv2d(256, 256, kernel_size=(3, 3), шаг=(2, 2), заполнение=(1, 1), смещение=False)
(bn2): FrozenBatchNorm2d(256, eps= 0,0)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn3): FrozenBatchNorm2d(1024, eps= 0.0)
(relu): ReLU(inplace=True)
(понижающая выборка): Sequential(
(0): Conv2d(512, 1024, kernel_size=(1, 1), шаг= (2, 2), смещение=False)
(1): FrozenBatchNorm2d(1024, eps=0.0)
)
)
(1): Узкое место(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn1): FrozenBatchNorm2d(256, eps=0.0)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), шаг=(1, 1), заполнение=(1, 1), смещение=False)
(bn2): FrozenBatchNorm2d(256, eps=0.0)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn3): FrozenBatchNorm2d(1024, eps=0.0)
(relu): ReLU(inplace=True)
)
(2): Узкое место (
(conv1): Conv2d(1024, 256, kernel_size=( 1, 1), шаг=(1, 1), смещение=False)
(bn1): FrozenBatchNorm2d(256, eps=0.0)
(conv2): Conv2d(256, 256, kernel_size=( 3, 3), шаг = (1, 1), заполнение = (1, 1), смещение = False)
(bn2): FrozenBatchNorm2d (256, eps = 0,0)
(conv3): Conv2d (256, 1024, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn3): FrozenBatchNorm2d(1024, eps=0.0)
(relu): ReLU (inplace=True)
)
(3): Узкое место (
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), шаг=(1, 1), смещение = False)
(bn1): FrozenBatchNorm2d (256, eps = 0,0)
(conv2): Conv2d (256, 256, kernel_size = (3, 3), шаг = (1, 1), дополнение = (1, 1), смещение = False)
(bn2): FrozenBatchNorm2d (256, eps = 0,0)
(conv3): Conv2d (256, 1024, kernel_size = (1, 1), шаг=(1, 1), смещение=False)
(bn3): FrozenBatchNorm2d(1024, eps=0.0)
(relu): ReLU(inplace=True)
)
(4): Узкое место(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn1): FrozenBatchNorm2d (256, eps=0,0)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), шаг=(1, 1), заполнение=(1, 1), смещение=False)< br /> (bn2): FrozenBatchNorm2d(256, eps=0,0)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), шаг=(1, 1), смещение=False)< br /> (bn3): FrozenBatchNorm2d(1024, eps=0.0)
(relu): ReLU(inplace=True)
)
(5): Узкое место(
(conv1 ): Conv2d(1024, 256, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn1): FrozenBatchNorm2d(256, eps=0.0)
(conv2 ): Conv2d(256, 256, kernel_size=(3, 3), шаг=(1, 1), заполнение=(1, 1), смещение=False)
(bn2): FrozenBatchNorm2d(256, eps= 0,0)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn3): FrozenBatchNorm2d(1024, eps= 0.0)
(relu): ReLU(inplace=True)
)
)
(layer4): Последовательный (
(0): Узкое место (
(conv1): Conv2d(1024, 512, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn1): FrozenBatchNorm2d(512, eps=0.0)
(conv2): Conv2d(512, 512, kernel_size=(3, 3), шаг=(2, 2), заполнение=(1, 1), смещение=False)
(bn2): FrozenBatchNorm2d(512, eps=0.0)
(conv3): Conv2d(512, 2048, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn3): FrozenBatchNorm2d(2048, eps=0.0)
(relu): ReLU(inplace=True)
(понижающая дискретизация): Sequential(
(0): Conv2d(1024, 2048, kernel_size=(1, 1), шаг=(2, 2), смещение=False)
(1): FrozenBatchNorm2d(2048, eps=0.0)
)
)
(1): Узкое место(
(conv1): Conv2d(2048, 512, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn1): FrozenBatchNorm2d(512, eps=0.0)
(conv2): Conv2d(512, 512, kernel_size=(3, 3), шаг=(1, 1), заполнение=(1, 1), смещение=False)
(bn2): FrozenBatchNorm2d(512, eps=0.0)
(conv3): Conv2d(512, 2048, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn3): FrozenBatchNorm2d(2048, eps=0.0)
(relu): ReLU(inplace=True)
)
(2): Узкое место (
(conv1): Conv2d(2048, 512) , kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn1): FrozenBatchNorm2d(512, eps=0.0)
(conv2): Conv2d(512, 512) , kernel_size=(3, 3), шаг=(1, 1), заполнение=(1, 1), смещение=False)
(bn2): FrozenBatchNorm2d(512, eps=0.0)
( conv3): Conv2d(512, 2048, kernel_size=(1, 1), шаг=(1, 1), смещение=False)
(bn3): FrozenBatchNorm2d(2048, eps=0.0)
( relu): ReLU(inplace=True)
)
)
)
(fpn): FeaturePyramidNetwork(
(inner_blocks): ModuleList(
(0 ): Conv2dNormActivation(
(0): Conv2d(256, 256, kernel_size=(1, 1), шаг=(1, 1))
)
(1): Conv2dNormActivation(< br /> (0): Conv2d(512, 256, kernel_size=(1, 1), шаг=(1, 1))
)
(2): Conv2dNormActivation(
(0 ): Conv2d(1024, 256, kernel_size=(1, 1), шаг=(1, 1))
)
(3): Conv2dNormActivation(
(0): Conv2d(2048 , 256, kernel_size=(1, 1), шаг=(1, 1))
)
)
(layer_blocks): ModuleList(
(0-3): 4 x Conv2dNormActivation(
(0): Conv2d(256, 256, kernel_size=(3, 3), шаг=(1, 1), дополнение=(1, 1))
)
)
(extra_blocks): LastLevelMaxPool()
)
)
(rpn): RegionProposalNetwork(
(anchor_generator): AnchorGenerator()
(head) : RPNHead(
(conv): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(256, 256, kernel_size=(3, 3), шаг=(1, 1), дополнение = (1, 1))
(1): ReLU(inplace=True)
)
)
(cls_logits): Conv2d(256, 3, kernel_size =(1, 1), шаг=(1, 1))
(bbox_pred): Conv2d(256, 12, kernel_size=(1, 1), шаг=(1, 1))
)
)
(roi_heads): RoIHeads(
(box_roi_pool): MultiScaleRoIAlign(featmap_names=['0', '1', '2', '3'], output_size=(7, 7), sample_ratio=2)
(box_head): TwoMLPHead(
(fc6): Linear(in_features=12544, out_features=1024, смещение=True)
(fc7): Linear(in_features =1024, out_features=1024, смещение=True)
)
(box_predictor): FastRCNNPredictor(
(cls_score): Linear(in_features=1024, out_features=6, смещение=True)
(bbox_pred): Линейный(in_features=1024, out_features=24, смещение=True)
)
)
)

Как преобразовать модель Pytroch в CoreML


Источник: https://stackoverflow.com/questions/780 ... -to-coreml