Convolutional neural network (CNN) in R | A Rstudio Tutorial on Keras and Tensorflow

[@itz_me_sick_gutz]

Informative and Detailed Video on KZbin about CNN ever. Thanks for sharing.

[@PleasantPixels]

Don't stop posting! Good teaching skill 👍

[@yaoli1413]

Well done, with great detail. Thanks for sharing.

[@phuwadejp4274]

Love you video. It makes thing much easier. Thank you.

[@shwetaredkar734]

Excellent video. Please make tutorial on graph convolution neural network for binary classification for non image data.

[@Casual_Wanderlust]

Lets say we want to load our own set of images into R how do we go about doing that? (really new to ML)

[@LiquidBrain]

I think this article summarized the process very well. This also seems to be a really good video, will try to do that soon. cran.r-project.org/web/packages/imager/vignettes/gettingstarted.html

[@gli6230]

Thank you very much, I am doing the homework on it.

[@ericoffice.0019]

thanks for sharing, quite clear and inspiring. by the way, isn't the right-lower convolution result 1.75? at ur spreadsheet table demo?

[@LiquidBrain]

Opps, something wrong with the copy paste i did. Thanks for pointing that out

[@mammerlin]

Great explanation video

[@tranquil2705]

I followed your script, but I encountered an error: Error in py_module_import(module, convert = convert): ModuleNotFoundError: No module named 'kerastools.callback'. Could you please help me?

[@nrcs91]

Same errorrr!!!

[@theforester_]

can you explain about the prediction? or recommend a video which talks about it?

[@LiquidBrain]

The prediction from this video is the type of image being fed in, is it a bus, aeroplane bicycle etc. Essentially, a image classifier?

[@theforester_]

@@LiquidBrain oh yep. Sorry i was wondering about the how to predict. Does it use predict() command? Or keras/tensorflow does prediction differently?

[@theforester_]

By the way i was wondering if i could feed satelite images in order to predict land cover classification. Have you ever tried it? Im a bit concerned about the size of the images which have got to have the same pixel counting right? Because some images have different resolutions which impact in the number of pixel...

[@LiquidBrain]

@@theforester_ Yeap, predict() is correct

[@EcoresolveInc]

Thanks a lot for the video!!!!

[@jennysilvestre123]

Hi! Can I use a numerical dataset for the CNN?

[@LiquidBrain]

Yap, you can ~

[@jennysilvestre123]

Do I need softwares other than RStudio such as miniconda from python? Because the keras sequential model says that there's no support interface from python so it needs to download miniconda.

[@LiquidBrain]

Keras and tensorflow actually only works in Python. What's R is doing here is using a conversation layer (call reticulate) to convert all the R code to Python so it can run on keras and tensorflow. So ya, you will.need to install miniconda(which is another program to help to handle Python and it's misc stuff) . Alternatively you can also run your CNN using “Torch for R”, it's slightly more complicated but it will run natively in R

[@carolinchensita]

@@LiquidBrain Thank you fot your video. I have been trying to work on this but for the past four hours I have been trying to install miniconda and get it to work, but it is not working... I have searched and viewed many videos and searched on forums but now I am very confused... Is there a video you could recommend on these steps?

[@LiquidBrain]

@@carolinchensita Installing Miniconda has also been complicated :). Can you check if you have the latest version of R and R studio? Video could be helpful and there's actually one I made previously, but with how much the package changes its always hard to know which one will work or not.