Skip to content
This repository has been archived by the owner on Jan 18, 2024. It is now read-only.

Problem with Yolo V3 #108

Open
jessiffmm opened this issue Feb 8, 2019 · 9 comments
Open

Problem with Yolo V3 #108

jessiffmm opened this issue Feb 8, 2019 · 9 comments

Comments

@jessiffmm
Copy link
Contributor

Hi!!

I'm triying to a yolov3 config and I get some errors. It said that it can't recognized yolo layer, upsampled layer,...
I think that https://github.com/JdeRobot/darknet/ isn't compatible with yolo v3

Regards

@vinay0410
Copy link
Collaborator

vinay0410 commented Feb 8, 2019 via email

@jessiffmm
Copy link
Contributor Author

82 Type not recognized: [yolo]
Unused field: 'mask = 6,7,8'
Unused field: 'anchors = 10,13,16,30,33,23,30,61,62,45,59,119,116,90,156,198,373,326'
Unused field: 'classes = 80'
Unused field: 'num = 9'
Unused field: 'jitter = .3'
Unused field: 'ignore_thresh = .7'
Unused field: 'truth_thresh = 1'
Unused field: 'random = 1'
83 route 79
84 conv 256 1 x 1 / 1 19 x 19 x 512 -> 19 x 19 x 256
85 Type not recognized: [upsample]
Unused field: 'stride = 2'
86 route 85 61
87 netwrigth /opt/datasets/weights//yolov3.weights./DatasetEvaluationApp: Violación de segmento (`core' generado)

@jessiffmm
Copy link
Contributor Author

I'm triying to introduce this type of layer in the detectionsuite.

@vinay0410
Copy link
Collaborator

vinay0410 commented Feb 10, 2019

I just executed yolov3, and it's not working with JdeRobot/darknet and with DetectionSuite. It's because we are stuck on an outdated fork of darknet, which doesn't support yolov3.

Although, Yolo v1 works well.

If Yolov3 is important for the task you are trying to accomplish, then we could shift to using Yolo's original repository instead of our own fork, which would fix this.

Just let me know, if you'd want that.

@jessiffmm
Copy link
Contributor Author

I have downloaded the last https://github.com/pjreddie/darknet and I have adapted it to JdeRobot/darknet for using DarknetApi and install in the computer.

I have this code in my repository (https://github.com/RoboticsURJC-students/2018-tfm-Jessica-Fernandez/tree/master/darknet).

You can used it if you want.

@vinay0410
Copy link
Collaborator

vinay0410 commented Feb 10, 2019

Okay, I will look into it tomorrow, and try to integrate it with DetectionSuite.
Although, if you are interested in learning more about DetectionSuite, you can submit a pull integrating this.

This file is supposed to be changed.

@jessiffmm
Copy link
Contributor Author

No, I adapt it for DetectionSuite. I've tried it with detectionsuite and it works perfectly. Only you have to use my darknet instead of https://github.com/JdeRobot/darknet/ .

I do not know how you could substitute https://github.com/JdeRobot/darknet/ for mine. I don't know if I have permissions.

@vinay0410
Copy link
Collaborator

vinay0410 commented Feb 10, 2019

Okay, great.
So, I will substitute your fork.

But, I was getting the following error while building it.

In file included from /home/robo/dev/2018-tfm-Jessica-Fernandez/darknet/DarknetAPI/DarknetAPI.cpp:5:0:
/home/robo/dev/2018-tfm-Jessica-Fernandez/darknet/DarknetAPI/DarknetAPI.h:12:10: fatal error: yolo.h: No such file or directory
 #include "yolo.h"
          ^~~~~~~~
compilation terminated.

This file isn't present anywhere in the repository.
Could you please fix this

Thanks !!

Once it builds we can substitute it.

@jessiffmm
Copy link
Contributor Author

Ohh sorry! Already solved!

Thank you!!

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants