search

Pongpisit-Thanasutives / Variations-of-SFANet-for-Crowd-Counting

The official implementation of "Encoder-Decoder Based Convolutional Neural Networks with Multi-Scale-Aware Modules for Crowd Counting"
https://ieeexplore.ieee.org/document/9413286
GNU General Public License v3.0
110 stars 32 forks source link

Session crashed after using all available RAM #26

Closed AminaZarief closed 2 years ago

AminaZarief commented 2 years ago

Hello, Thanks for your greatest effort, it really helps me a lot, but I have an issue when testing on the image(img_0071.jpg), the question here, Is the suggested model take more than 13GB RAM?, I have already followed the instructions suggested in the example code, but it gave me "Session crashed after using all available RAM"

Pongpisit-Thanasutives commented 2 years ago

@AminaZarief I am sorry for such high use of computational resources. I tested on Google Colab's GPU and got a similar result. For now, a simple solution could be dividing the target image into 4 equal patches and then estimating the counts for each of them.

Thank you for the issue, I might also change the example image since it's quite large in resolution.

AminaZarief commented 2 years ago

@AminaZarief I am sorry for such high use of computational resources. I tested on Google Colab's GPU and got a similar result. For now, a simple solution could be dividing the target image into 4 equal patches and then estimating the counts for each of them.

Thank you for the issue, I might also change the example image since it's quite large in resolution.

Thank you for your reply, I will give it a try