architectureIntroduced by GoogLeNet (Inception) · 2014
Inception module
Parallel convolutions of several sizes in one block, to go deeper and wider at fixed cost.
Drafted by AI · not yet reviewed
How this idea evolved
Suggested from citations. No curator has recorded what this idea builds on yet. These ideas from the same theme come from papers that GoogLeNet (Inception) cites, directly or one step removed. Citation is a fact; the connection between the ideas is not verified.
Project activations back to pixel space to see what each CNN layer has learned.
Papers using this
- 2014VGG
- 2015PReLU / He init
- 2015BatchNorm
- 2015YFCC100M
- 2015Ask Your Neurons
- 2015FM-IQA
- 2015Inception v3
- 2016Wide ResNet
- 2016ResNeXt
- 2018ImageNet-trained CNNs are biased towards
- 2019EfficientNet
- 2021ViT-VQGAN