YOLOP

YOLOP copied to clipboard

我看网络模型在语义分割部分输出时，添加了nn.Sigmoid函数，但是在计算损失函数的时候有用了nn.BCEWithLogitsLoss，这个函数内部不是又做了一次sigmoid操作么，这个会不会存在问题

qq751220449

training error

2

Traceback (most recent call last): File "train.py", line 395, in main() File "train.py", line 322, in main train(cfg, train_loader, model, criterion, optimizer, scaler, File "/home/zhoup/work/download_source_code/yolop/lib/core/function.py", line 77, in train total_loss,...

the-cat-crying

矩形输入问题

1

你好，部署模型的时候一般都使用矩形图，而不是方形图。我在作多任务训练测试时，发现padding图片到方形，训练正常，用训好的模型在不padding进行矩形输入推理时，车道线指标存在严重下降的现象。当我使用矩形训练时，车道线指标上不去或者随着训练epoch增加车道线指标越来越低的现象。不知道作者有没遇到这个问题？

chzhan

Hello, thank you for the open source project. I have a question, Can we merge the two segmentation heads into just one head? Thanks

Khaliladib11

Great job with this repo! I would to ask what you believe that is the best procedure to follow in order to make my **1:10 scaled vehicle** detect lines on...

TheoPapafotiou

Training model with only drivable area segmentation

4

I am trying to train the model with only one task - drivable area segmentation. I have been trained on the machine with GTX 1080Ti for three days already. Now...

danielzhangau

Recieved error related to performing NMS while training on Colab

2

I am training BDD100K driveable segmentation area identification task on colab However after one epoch of training is completed during validation I end up recieving the following error ``` 0%...

sparshgarg23

raimannma

Does anyone know how to obtain only the lines or only the roads from the semantic segmentation images? Because I wanted to generate a database with the images of each...

jonathanmg96

docker run RuntimeError:Dataloader worker (pid 6479) is killed by signal:killed

the-cat-crying