{
  "id": 162064,
  "title": "How can we import a compiled .so package of Yolov4?",
  "url": "/competitions/global-wheat-detection/discussion/162064",
  "author_name": "",
  "post_date": "2020-06-27T08:02:24.358000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi Kagglers,\n  I recently tried a YoloV4 model, which is implemented in pure C. I make the model to create a <code>libdarknet.so</code> to initiate the model. However, Kaggle kernel seems does not allow me to use that. If the <code>.so</code>  file was compiled with CUDA &amp; OpenCV, it's can't be loaded successfully on Kaggle kernels because we can't access <code>libcuda.so</code> or <code>libopencv-2.***.so</code> stuff. \n  So my current embarrassing situation is that I got a well-performed model but I can't submit...which is quite annoying.</p>\n\n<p>Anyone has experience dealing with such kind of models on Kaggle?</p>\n\n<p>Thank you in advance!</p>",
  "messages": [
    {
      "id": 903917,
      "postDate": "2020-06-27T08:02:24.360Z",
      "content": "<p>Hi Kagglers,\n  I recently tried a YoloV4 model, which is implemented in pure C. I make the model to create a <code>libdarknet.so</code> to initiate the model. However, Kaggle kernel seems does not allow me to use that. If the <code>.so</code>  file was compiled with CUDA &amp; OpenCV, it's can't be loaded successfully on Kaggle kernels because we can't access <code>libcuda.so</code> or <code>libopencv-2.***.so</code> stuff. \n  So my current embarrassing situation is that I got a well-performed model but I can't submit...which is quite annoying.</p>\n\n<p>Anyone has experience dealing with such kind of models on Kaggle?</p>\n\n<p>Thank you in advance!</p>",
      "rawMarkdown": "Hi Kagglers,\n  I recently tried a YoloV4 model, which is implemented in pure C. I make the model to create a `libdarknet.so` to initiate the model. However, Kaggle kernel seems does not allow me to use that. If the `.so`  file was compiled with CUDA &amp; OpenCV, it's can't be loaded successfully on Kaggle kernels because we can't access `libcuda.so` or `libopencv-2.***.so` stuff. \n  So my current embarrassing situation is that I got a well-performed model but I can't submit...which is quite annoying.\n\n  Anyone has experience dealing with such kind of models on Kaggle?\n\n  Thank you in advance!"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "903917": "Hi Kagglers,\n  I recently tried a YoloV4 model, which is implemented in pure C. I make the model to create a `libdarknet.so` to initiate the model. However, Kaggle kernel seems does not allow me to use that. If the `.so`  file was compiled with CUDA &amp; OpenCV, it's can't be loaded successfully on Kaggle kernels because we can't access `libcuda.so` or `libopencv-2.***.so` stuff. \n  So my current embarrassing situation is that I got a well-performed model but I can't submit...which is quite annoying.\n\n  Anyone has experience dealing with such kind of models on Kaggle?\n\n  Thank you in advance!"
  }
}