I tried to extract features from following code. However, it says 'FasterRCNN' object has no attribute 'features' I want to extract features with (36, 2048) shape features when it has 36 classes. Is there any method to extract with pretrained pytorch models.
model = torchvision.models.detection.fasterrcnn_resnet50_fpn(pretrained=True).to(device)
features = list(model.features)
dummy_img = torch.zeros((1, 3, 800, 800)).float() # test image array
req_features = []
output = dummy_img.clone().to(device)
for feature in features:
output = feature(output)
if output.size()[2] < 800//16: # 800/16=50
break
req_features.append(feature)
out_channels = output.size()[1]
faster_rcnn_feature_extractor = nn.Sequential(*req_features)
output_map = faster_rcnn_feature_extractor(dummy_img )
print(output_map.shape)
The function you are calling returns a FasterRCNN
object which is based on GeneralizedRCNN
. As you have experienced, this object doesn't indeed have a feature
attribute. Looking at its source code, if you want to acquire the feature maps, you can follow L83 and L101 :
>>> images, _= model.transform(images, None)
>>> features = model.backbone(images.tensors)
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.