Im trying to get an output from a pretrained resnet_fpn_backbone for processing in object detection. Im using the finetuning model from torchvision (fasterrcnn_resnet50_fpn) to test pretraining. Unfortunately i failed several times and i don’t know what to change.
I used the following code (left out imports):
#MAKE MODEL
class Identity(nn.Module):
def __init__(self):
super(Identity, self).__init__()
def forward(self, x):
return x
model = torchvision.models.detection.fasterrcnn_resnet50_fpn(pretrained=True)
model.rpn = Identity()
model.roi_heads = Identity()
#MAKE INPUT-TEST-DATA
images = torch.rand(2, 3, 600, 1200)
boxes = np.array([[[0.29, 0.10, 0.43, 0.58],[0.24, 0.17, 0.33, 0.48]],[[0.04, 0.10, 0.63, 0.58],[0.29, 0.16, 0.43, 0.78]]])
boxes = boxes.astype(np.float32)
boxes = torch.from_numpy(boxes)
labels = torch.randint(1, 91, (2, 2))
images = list(image for image in images)
targets = []
for i in range(len(images)):
d = {}
d['boxes'] = boxes[i]
d['labels'] = labels[i]
targets.append(d)
#TEST MODEL (in model.train())
output = model(images, targets) #I fail here!
print(output)
This is the error i receive:
