Byronnar
Byronnar
你用的代码就是上面那个? 我concat之后 ,可以在serving端部署.
> > 这个我理解的就是作者已经使用类似这种代码将yolo3.weight文件转换成了ckpt，也就是说直接使用coco.ckpt权重就可以了，所以就直接把转的那步省略了，但是因为大家都是对yolo3.weight文件在意的比较多，所以产生了误导吧 > > 我猜测也是这样，只是想与作者确认一下，如果我自己将darknet的weight转化为ckpt，使用他其他的code有没有什么需要注意的地方。比如，作者将yolov3原始的anchor都scale了一下，类似于这种小细节。 请问你可以用这个权重训练自己的数据集嘛？我改了.names文件之后一直无法载入预训练模型，路径正确，./checkpoint/yolov3_coco_demo.ckpt does not exist !!! 期待您的回复
> when I try python build.py build_ext --inplace I met error as follows: > >  > could anybody give me some advice??thank you very much and thank for author`s...
> mask_batch shape: torch.Size([1, 3, 320, 320]) > This means your masks have 3 channels, i.e., three segmentation classes. I hard-coded the polyp task to have 2 classes (according to...
> mask_batch shape: torch.Size([1, 3, 320, 320]) > This means your masks have 3 channels, i.e., three segmentation classes. I hard-coded the polyp task to have 2 classes (according to...
> Not sure what you mean by inference. test2d.py and test3d.py do inference on 2d and 3d datasets. Do you mean doing inference on a _single image_ instead of a...
> Well it's easy 😁 (but a bit hacky) you can put the test images in a subfolder under the task data folder. Then add a dataset to test2d.py and...
Mayby last question, I have trained a model, but hen I run test.py, there are some mistakes: python test2d.py --task polyp --split all --ds Kvasir-test --net segtran --translayers 3 --layercompress...
> Seems I should provide a command line argument to allow overriding the hardcoded number of classes 🤔 My dataset name is Kvasir-SEG, it is downloaded from https://datasets.simula.no/kvasir-seg/, number_classes should...
这是对应的输出层啊, 第一个是输入层