search

liulei01 / DRBox

A deep learning based algorithm to detect rotated object, for example, objects in remote sensing images
Other
421 stars 142 forks source link

out of memory #19

Open fragileness opened 5 years ago

fragileness commented 5 years ago

I tried to execute "python deploy.py" under "drbox/examples/rbox/deploy/Airplane", and the error is:

Check failed: error == cudaSuccess (2 vs. 0) out of memory

I'm using GeForce GTX 1070 (8GB). Does it means that GeForce GTX 1070 has not enough memory to run DRBox on it? What's the hardware requirement for DRBox?

keck91 commented 5 years ago

According to a Report of the Department of Computer Science, Virginia Tech in 2018, DRBox requires 11GB GPU memory.

Source (page 29)

EthanGreen75 commented 5 years ago

How about use smaller batch Size or smaller image?

Viktor-Paul commented 3 years ago

I tried to execute "python deploy.py" under "drbox/examples/rbox/deploy/Airplane", and the error is:

Check failed: error == cudaSuccess (2 vs. 0) out of memory

I'm using GeForce GTX 1070 (8GB). Does it means that GeForce GTX 1070 has not enough memory to run DRBox on it? What's the hardware requirement for DRBox?

Have you solved this problem?

Viktor-Paul commented 3 years ago

How about use smaller batch Size or smaller image?

I tried to use smaller batch Size or smaller image, but it still failed. Is it because my memory is too small?

Viktor-Paul commented 3 years ago

I'm very happy that I've solved the problem。In the deploy/Airplane/deploy. Prototxt file,I modified the first dim of shape in dummy_data_param to 1, because this parameter is the number of data augmentation for recognition samples, and I need to recognize the whole image without augmenting the image data.