Loading timeline…
20152010s
Residual Networks (ResNets)
An architecture using skip connections to enable the training of hundred-layer deep networks.
Why It Was Important
Before ResNets, adding more layers to a CNN actually worsened performance due to vanishing gradients. Kaiming He and his team at Microsoft Research solved this by adding 'skip connections' that let signals skip layers, enabling networks with 152+ layers to dominate computer vision tasks outlandishly.
Who Invented It
Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun
Computer vision researchers at Microsoft Research Asia.
Applications
- Computer Vision Models
- Feature Extraction Backbones
- Autonomous Driving
Key Papers
- Deep Residual Learning for Image Recognition
Kaiming He et al. · CVPR 2016 · 2015
Videos
C4W2L03 Resnets
DeepLearningAI
Residual Networks and Skip Connections (DL 15)
Professor Bryce