- Notifications
You must be signed in to change notification settings - Fork964
[FEAT] Add dilate and erode operations to Tensor#1136
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to ourterms of service andprivacy statement. We’ll occasionally send you account related emails.
Already on GitHub?Sign in to your account
base:main
Are you sure you want to change the base?
Uh oh!
There was an error while loading.Please reload this page.
Conversation
| // Only include values where the kernel has a value | ||
| // of 1. | ||
| // Rather than multiply against this value, we use | ||
| // the if check to reduce the size of the array. | ||
| const kernelValue = kernel[kernelRowOffset + paddingSize.height][kernelColOffset + paddingSize.width]; | ||
| if (kernelValue === 1) { | ||
| kernelValues.push(data[neighborIndex] * kernelValue); | ||
| } |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others.Learn more.
Originally I was not using the if statement, and utilising the fact that a kernel contains either0 or1, however that meant thatkernelValues array would be bigger than it needs to be.
Also,erode wouldn't work properly because0 is always the minimum value.
| function validateKernel(kernelSize) { | ||
| let kernel; | ||
| if (typeof kernelSize === 'object' && 'width' in kernelSize && 'height' in kernelSize) { | ||
| // This is a Size object, so no conversion required. | ||
| kernel = kernelSize; | ||
| } else if (typeof kernelSize === 'number' && Number.isInteger(kernelSize)) { | ||
| // A single whole number is assumed as the width and height. | ||
| kernel = { width: kernelSize, height: kernelSize }; | ||
| } else if (Array.isArray(kernelSize) && kernelSize.length === 2 && kernelSize.every(Number.isInteger)) { | ||
| // An array of two values is assumed as width then height. | ||
| kernel = { width: kernelSize[0], height: kernelSize[1] }; | ||
| } else { | ||
| throw new Error("Invalid kernel size."); | ||
| } | ||
| if (kernel.width % 2 === 0 || kernel.height % 2 === 0) { | ||
| throw new Error("Kernel size must be odd"); | ||
| } | ||
| return kernel; | ||
| } |
BritishWerewolfJan 4, 2025 • edited
Loading Uh oh!
There was an error while loading.Please reload this page.
edited
Uh oh!
There was an error while loading.Please reload this page.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others.Learn more.
The purpose of this function is to help users with passing in kernel values.
To get a 3x3 kernel to theerode function we can do this:
tensor.erode(3)tensor.erode([3, 3])tensor.erode({ width: 3, height: 3 })
Most likely we want a symmetrical kernel, and it's easier to just pass in a single number.
a1a2ce8 tocfdd3b2Comparecfdd3b2 to22a551cCompare
Uh oh!
There was an error while loading.Please reload this page.
This PR is inspired by the functionality ofOpenCV.
With this, we can create a kernel that is operated over each element of the
dataarray and performs some calculation.For this PR, we only have
dilateanderode.I decided to set these functions to
async, despite that makingthe chaining a bit clunkier, since these functions are often used together.Explanation
kernelThis is an array of 0s and 1s that will be applied to each element of the
dataarray, 1 by 1.There are multiple shapes supported in this PR, they are:
RECT,CROSS,ELLIPSE.Consider a 5x5 kernel for each shape:
RECT:CROSS:ELLIPSE:The element we are checking will be in the centre, and then it will apply this kernel to all those values around it.
Any value of 1 will be included and 0 will be discarded.
dilateOnce we have a list of values for that element, we will apply a
Math.maxto return the biggest value of these.Consider this Tensor.
If we were to apply
CROSSwith a kernel size of 3x3, then the first iteration would return the values of[1, 2, 6]. This is because the top and left side of the kernel are off the edge.When the loop gets to the value of
13, the extracted values would be[8, 12, 13, 14, 18].Finally, the value in each instance will be replaced with the maximum value.
Again, in the case of the first element, that would be
6and for the centre element, that would be replaced with18.erodeThis works in the exact same way as
dilate, however it will apply theMath.minto the elements instead.