BiSeNet

BiSeNet copied to clipboard

TRT inference for two images does not work properly

2

Hi! - Trained BisenetV2 model and converted it .trt model successfully - When we infer on one image model shows correct result in infer - but when we send 2...

syedrz

Training errow

2

 How to solve this problem？

Jasonx0123

use for kitti, how to train or finetune,thanks!

2

Dear author, If I want to use the model to do the training in the Kitti dataset. So we have the different category number, for the sityscapes ,it is 20...

emilyemliyM

Hi @CoinCheung , I am using my own custom dataset to train bisenetv2 model. Dataset Details : image size : 1920x1080 Only two values in my label image {0, 127}...

shilpaullas4

windows下应该怎么训练？

1

C:\Users\97328\.conda\envs\torch\python.exe "E:\A_Deep learning model\BiSeNet\tools\train_amp.py" Traceback (most recent call last): File "E:\A_Deep learning model\BiSeNet\tools\train_amp.py", line 210, in main() File "E:\A_Deep learning model\BiSeNet\tools\train_amp.py", line 200, in main local_rank = int(os.environ['LOCAL_RANK']) File "C:\Users\97328\.conda\envs\torch\lib\os.py",...

t973288913

As we know, ade20k has 150 categories (1-150). But in the configuration file, does n_cats=150 mean that background 0 is not included? My question is actually when we do binary...

genhao3

prepare dataset的ade20k中命令缺了一块 `$ python tools/gen_dataset_annos.py --ade20k` `$ python tools/gen_dataset_annos.py --dataset ade20k`

peterhuang621

ValueError: Error initializing torch.distributed using tcp:// rendezvous: rank parameter missing 

Lv595

关于tensorrt加速的问题

1

我是在nano板上进行的模型部署的，但是由于nano板自带tensor7.1.3的版本，我无法部署您在代码中采用的tensorrt加速代码，于是我采用了一个适配我的nano板的加速代码，但最终得到的效果很差，请问能不能提供一些解决方法或如何修改您的加速代码使得能在nano板上tensorrt7.1.3版本运行，谢谢。我采用您的tensorrt加速代码在PC端进行的分割效果图，比在nano板上运行的效果要好很多，所以我猜测是这个原因导致的。

Joy2Boy

关于config文件里的n_cats设置

10

我通过check_dataset_info.py脚本查到我的所有数据集标签图像像素值区间是[0，132]，但是我的数据集标签类别一共才13类，为什么n_cats设成13我的训练就会报错呢，但是设成133又能成功运行？ 

yangaiping