Very Deep Convolutional Networks for Large-Scale Image Recognition
Citations
491 citations
Cites background or methods from "Very Deep Convolutional Networks fo..."
...DNNs have dramatically improved the state of the art in many challenging problems (13), including speech recognition (20–22), machine translation (23, 24), image recognition (25, 26), and playing Atari games (27)....
[...]
...neurons) and obtains better performance than AlexNet by using effective 3 × 3 convolutional filters (26)....
[...]
489 citations
489 citations
487 citations
Cites methods from "Very Deep Convolutional Networks fo..."
...To simplify the description, all subsequent model parameters are based on the VGG-16 backbone....
[...]
...Specifically, we remove the last max-pooling layer of the VGG-16 to maintain the details of the final convolutional layer....
[...]
...Our model is built on the FCN architecture with the pretrained VGG-16 [29] or ResNet-50 [12] as the backbone, both of which only retain the feature extraction network....
[...]
...The backbone parameters (i.e. VGG-16 and ResNet-50) are initialized with the corresponding models pretrained on the ImageNet dataset and the rest ones are initialized by the default setting of PyTorch....
[...]
...Our model takes a RGB image (320×320×3) as input, and exploits VGG-16 [29] blocks {E}(4)i=0 to extract multi-level features....
[...]
486 citations
References
49,639 citations
21,729 citations
9,803 citations
9,775 citations
6,061 citations