Deep Residual Learning for Image Recognition
GPTKB entity
Statements (48)
| Predicate | Object |
|---|---|
| gptkbp:instance_of |
gptkb:academic_journals
|
| gptkbp:application |
Image Classification
|
| gptkbp:architecture |
gptkb:Convolutional_Neural_Network_(CNN)
|
| gptkbp:author |
gptkb:Shaoqing_Ren
gptkb:Kaiming_He gptkb:Xiangyu_Zhang Jian Sun |
| gptkbp:challenges |
Utilizes skip connections for better feature propagation
Facilitates training of networks with hundreds of layers Achieves better performance with fewer parameters Deeper networks can be trained effectively Demonstrated effectiveness on various benchmarks Encourages exploration of new architectures in deep learning Improved accuracy with increased depth Introduced the concept of residual blocks Residual connections mitigate vanishing gradient problem Promotes the use of deeper architectures in practice |
| gptkbp:contribution |
Advancement of deep learning techniques
Introduction of residual networks (Res Nets) |
| gptkbp:feature |
gptkb:Batch_Normalization
gptkb:Dropout Data Augmentation Weight Initialization Non-linear Activation Functions Gradient Descent Optimization Layer-wise training |
| https://www.w3.org/2000/01/rdf-schema#label |
Deep Residual Learning for Image Recognition
|
| gptkbp:impact |
Improved training of deep neural networks
|
| gptkbp:influence |
Inspired further research in deep learning architectures
|
| gptkbp:is_cited_in |
Highly cited
|
| gptkbp:key_concept |
Identity Mapping
Skip Connections Residual Learning |
| gptkbp:performance |
Top-5 Error Rate
|
| gptkbp:provides_information_on |
gptkb:Image_Net
|
| gptkbp:published_in |
gptkb:Proceedings_of_the_IEEE_Conference_on_Computer_Vision_and_Pattern_Recognition_(CVPR)
|
| gptkbp:published_year |
gptkb:2016
|
| gptkbp:related_to |
gptkb:Computer_Vision
gptkb:Artificial_Intelligence gptkb:Deep_Learning |
| gptkbp:research |
Deep Learning Techniques
|
| gptkbp:result |
Achieved state-of-the-art performance on Image Net
|
| gptkbp:subsequent_work |
gptkb:Dense_Net
gptkb:Res_Ne_Xt gptkb:Res_Net_Variants gptkb:Wide_Res_Net |
| gptkbp:bfsParent |
gptkb:Ilya_Sutskever
|
| gptkbp:bfsLayer |
4
|