I'm trying to reproduce this tutorial to make sure I understand how things work before training on my data, so I am using his dataset and labels.
I am not very sure how to structure the files so I guess the issue has something to do with it.
I am training on an AWS EC2 server with CUDA 8 and OpenCV 3 pre-installed.
main directory darknet
I am guessing obj.data is the culprit
classes= 1
train = data/nfpa/train.txt
valid = data/nfpa/test.txt
names = data/obj.names
backup = backup/
Part of my Makefile
GPU=1
CUDA=1
CUDNN=1
OPENCV=1
OPENMP=0
DEBUG=0
My issue is 
I have checked the text files and the values are normal (not infinity).
Please help. Thank you!
From the screenshot it seems like you have labels (.txt) and images in same folder it should be in the form of:
train.txt should contain paths to:
and then your labels should be placed and have the structure like this:
To be able to provide better help I would like the following:
Sorry, I forgot to update. I downloaded the VOC dataset and set it up as instructed in the README and things are working now. Before, I got my images and labels folders all mixed up. Thank you for your help :)
check out my comment: https://github.com/pjreddie/darknet/issues/174#issuecomment-445203621
It doesn't have to do with the CUDA or cuDNN versions.
I am getting the same issue Cannot load image "./darknet/dataset/z1.jpg" STB Reason: can't fopen Please help me with this @antnh6 @sachindesh @TheMikeyR
i have same issue:
images *.jpg convert into *.png format.
Fixed.
i have same issue:
images *.jpg convert into *.png format.
Fixed.
I have the same problem, and it worked when I used a png format, but why this happened
Most helpful comment
i have same issue:
images *.jpg convert into *.png format.
Fixed.