What does ResNet-50 mean?

A convolutional neural network that is 50 layers deep.

What are residual blocks?

Residual blocks are skip-connection blocks that learn residual functions with reference to the layer inputs. Each residual block contains 3 layers. There are 16 (3+4+6+3) residual blocks in ResNet-50 architectures.

What is the purpose of 1x1 convolutions?

1x1 convolutions are used to shrink down the representation significantly without hurting the performance.

Where is Batch Normalization (BN) used?

BN is used right after each convolution and before activation.

What is the bottleneck architecture?

According to the ResNet paper (pg. 6), The three layers are 1×1, 3×3, and 1×1 convolutions, where the 1×1 layers are responsible for reducing and then increasing restoring dimensions, leaving the 3×3 layer a bottleneck with smaller input output dimension. Both designs share the same time complexity.