Docker not able to load model for inference using dll invoke

[nielsmunters]

For a project I am using Yolov4 as a object detection model in a c# application inside of a docker container.

The problem: When trying to load the model the application crashes with different error codes. The compiled darknet works on the container but there seems to be a problem with the generated libdarknet.so file.

Errors: Error1:

Error2:

The setup: My system: Intel laptop with 2070 MaxQ - Windows OS Inside of a Docker container I built Darknet with GPU=1, CUDNN=1, CUDNN_HALF=1, OPENCV=1 and LIBSSO=1. Architecture was set to "-gencode arch=compute_75,code=[sm_75,compute_75]". The Docker container was based on: nvidia/cuda:11.4.2-cudnn8-devel-ubuntu20.04

In the c# application I try to load the model using:

Is this an issue with Darknet or with Docker? Any help would be greatly appreciated!

UPDATE: Issue was fixed by using older commit. Commit used was: https://github.com/AlexeyAB/darknet/commit/08088dccbbae7bb68c41d5ae3355380ce6654663

[nielsmunters]

Issue was fixed by using older commit. Commit used was: 08088dccbbae7bb68c41d5ae3355380ce6654663

[mhwahdan]

I beleive the problem is with docker.

you can try this fork of the repository

RoBen-AAST/darknet

All you need is to run: docker-compose run yolo-gpu bash you can also adjust the docker-compos file to suit your needs and don't forget to configure the Makefile

[nielsmunters]

I beleive the problem is with docker.

you can try this fork of the repository

RoBen-AAST/darknet

All you need is to run: docker-compose run yolo-gpu bash you can also adjust the docker-compos file to suit your needs and don't forget to configure the Makefile

I've fixed the error by using an older version of Darknet, so not sure it is just an issue with docker. But thanks for your comment!