Hi, may be my question is a bit unrelated (being very specific to implementation part), but I am facing problems in running my deep learning method on the validation hazy images. The image size is around 4000x3000 and in the 'Evaluation' section of competition details, it is mentioned that we have to submit the high resolution dehazed images. I am always getting out of memory error when I run my model on the input image of this size. I have a Nvidia Titan X GPU, is there a trick like splitting image into sub-images? Thank you!
Posted by: kunal @ Feb. 6, 2018, 4:22 p.m.Run it on CPU after training. I guess it will work.
Posted by: bchanda @ Feb. 7, 2018, 1:13 p.m.