Commit 7ba12d6

authored

Update readme, add images

1 parent b96058a commit 7ba12d6Copy full SHA for 7ba12d6

File tree

-1

lines changed

-1

lines changed

Lines changed: 9 additions & 1 deletion

Original file line number	Diff line number	Diff line change
`@@ -12,6 +12,10 @@ Capsules are a small group of neurons that have a few key traits:`
`12`	`12`	`* Every capsule outputs a vector, which has some magnitude and orientation.`
`13`	`13`	`* Capsules have a hierarchy between child and parent capsules and use dynamic routing to find the strongest connections between the output of one capsule and the inputs of the next layer of capsules.`
`14`	`14`
	`15`	`+<p align="center" >`
	`16`	`+ <img src='./assets/cat_face_2.png' width=60% />`
	`17`	`+</p>`
	`18`	`+`
`15`	`19`	`You can read more about all of these traits in [my blog post about capsules and dynamic routing](https://cezannec.github.io/Capsule_Networks/).`
`16`	`20`
`17`	`21`	`### Representing Relationships Between Parts`
`@@ -28,4 +32,8 @@ The Capsule Network that I'll define is made of two main parts:`
`28`	`32`	`1. A convolutional encoder`
`29`	`33`	`2. A fully-connected, linear decoder`
`30`	`34`
`31`		`-The notebook follows the architecture described [in the original Capsule Network paper](https://arxiv.org/pdf/1710.09829.pdf).`
	`35`	`+<p align="center" >`
	`36`	`+ <img src='./assets/complete_caps_net.png' width=80% />`
	`37`	`+</p>`
	`38`	`+`
	`39`	`+The above image was taken from the original [Capsule Network paper (Hinton et. al.)](https://arxiv.org/pdf/1710.09829.pdf). The notebook follows the architecture described in that paper and tries to replicate some of the experiments, such as feature visualization, that the authors pursued.`

Comments

(0)