search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.65k stars 7.96k forks source link

Segmentation fault (core dumped) in training own data yolo2 darknet #254

Open aminullah6264 opened 6 years ago

aminullah6264 commented 6 years ago

i have prepare my data according to the instruction given in this link

https://pjreddie.com/darknet/yolo/

i have downloaded weights also. i have 12GB titanX gpu when i run the darknet for training it give this error. i use following line for training ./darknet detector train cfg/voc.data cfg/yolo-voc.cfg darknet19_448.conv.23

plz how to solve this problem or why it is comming

image

AlexeyAB commented 6 years ago

@Aminullah6264 Hi,

  1. Did you use this repo or original repo?
  2. Did you get this error even for ./darknet detector test ...?
  3. What GPU, CUDNN, OPENCV, ... did you use in your Makefile?
  4. Try to compile with GPU=0, CUDNN=0, OPENCV=0, OPENMP=0 in the Makefile, so can train/test works successfully?
aminullah6264 commented 6 years ago

Dear @AlexeyAB
when i run ./darknet detector test it works for pretrained model but not working for example data training

AlexeyAB commented 6 years ago

@Aminullah6264 Try to use this release: https://github.com/AlexeyAB/darknet/releases It tested on both Windows and Linux for training and detection.

So if it doesn't work for your, then your problem in your environment / libraries (OpenCV, CUDA, cuDNN, ...)

xitman47 commented 6 years ago

@AlexeyAB hi!

Unfortunately, I have the same issue on Ubuntu 16. I will try to use earlier release as suggested. Here is the log:

Loading weights from yolo.weights... seen 32 Done! Learning Rate: 0.001, Momentum: 0.9, Decay: 0.0005 eval: Using default 'voc' 4 "annot load image "data/dog.jpg Segmentation fault (core dumped)

xitman47 commented 6 years ago

@AlexeyAB unfortunately, the earlier release does not work either. To be fair what I do now is a bit different from the topic's problem but I decided to put a note in this thread as I guess issues are still similar. My task is to get a list of coordinates for images in folder using valid command on pretrained coco weights.

I googled issue and I thought it is due to not having proper rights on AWS server but changing them did not help. And I am not sure this is the case also cause test does work and save predictions. I tried changing images names in the .txt file in many ways but YOLO still cannot load them:

Update: predictions are empty in the results folder when running test.

Loading weights from yolo.weights... seen 32 Done! Learning Rate: 0.001, Momentum: 0.9, Decay: 0.0005 eval: Using default 'voc' 4 "annot load image "data\dog.jpg Cannot load image "(null)" Cannot load image "data\eagle.jpg" sh: 1: Syntax error: word unexpected (expecting ")") Segmentation fault (core dumped)