Hello, why do you use StratifiedKFold loop this way? Normally you need to reset the model in each iteration of the for-loop. Your model is incrementally changed in each iteration, therefore it is not using validation data as un-seen data. It has been seen in the previous iteration ! Therefore the measurement of its performance is not correct. When you reset the model correctly then you get this result: fp= 2 fn= 491 tp= 14 tn= 993 Also, you have a testing directory with images 1500-2057, but no labels for it. Why ? Have you ever run the model on really un-seen data in "testing" directory ?
Yes, this is absolutely correct!! @vengelsno77. I should have used build_cnn() within the for loop and not outside!! Also, this data from where we have gathered was in this format. There were no test labels provided. If I am not wrong, it was a competition dataset as it also had submission.csv in the main dataset.
I've cross checked it, the data link is working fine. Please reach out to us on Instagram if you are unable to download the dataset (@spartifical_ is our insta handle)
Thanks for sharing, just wondering how to have access to live geo spatial data? Suppose I want to test my model on live dataset.
What exactly do you mean by live here? Data from Google Map API?
@@rohanshah8129 yes google maps ... can you tell us how can we do that please?
Hello! excellent tutorial. I did not find the dataset of images in the link, can you please make it available? thank you
Sorry, I already found hehehe. Thank you
Thank you 💯
Thank you bro! ✨
Hello, why do you use StratifiedKFold loop this way? Normally
you need to reset the model in each iteration of the for-loop.
Your model is incrementally changed in each iteration, therefore
it is not using validation data as un-seen data.
It has been seen in the previous iteration !
Therefore the measurement of its performance is not correct.
When you reset the model correctly then you get this result:
fp= 2
fn= 491
tp= 14
tn= 993
Also, you have a testing directory with images 1500-2057,
but no labels for it. Why ?
Have you ever run the model on really un-seen data in "testing" directory ?
Yes, this is absolutely correct!! @vengelsno77. I should have used build_cnn() within the for loop and not outside!!
Also, this data from where we have gathered was in this format. There were no test labels provided. If I am not wrong, it was a competition dataset as it also had submission.csv in the main dataset.
Thanks for this video, would you please give us the dataset link ?
I have changed the link in the notebook. Please Check it again. Thanks for pointing it out.
Could you reupload the dataset ? Anyway nice Video it helped me a lot !
It is done @nicolashoffmann8805
Can you provide the reference paper/conference paper/journal paper for it
Paper for the solar panel detection using ML?
dataset is not available, please modify the google drive link, thanks
I've cross checked it, the data link is working fine. Please reach out to us on Instagram if you are unable to download the dataset (@spartifical_ is our insta handle)
@@Spartificial thank you so much for your kind reply sir,
@@Spartificial i am going to apply transfer learning on this dataset, can you share the paper of this task?
I did not find the dataset of images
It is mentioned in the notebook itself.
drive.google.com/drive/folders/1Pe887pD6DP0nsqq0G1BdoMNExDvBoSp9
Can you please provide a link of colab
github.com/Spartificial/yt-acad-projs/blob/main/rspd-1222.ipynb
The drive link is not working its 404
I have changed the link in the notebook. Please Check it again. Thanks for pointing it out.
the data not clear
As in you didn't understand the data or the quality is not good?
@@rohanshah8129 no its just the link of the data dont work for me
I have changed the link in the notebook. Please Check it again. Thanks for pointing it out.