search

hellochick / ICNet-tensorflow

TensorFlow-based implementation of "ICNet for Real-Time Semantic Segmentation on High-Resolution Images".
405 stars 153 forks source link

tensorflow's version #120

Open leedoge opened 4 years ago

leedoge commented 4 years ago

thank you for providing the useful code! but when i run the demo file there are some error about the tensorflow's version and about the cuda cudnn. could you please tell me the reason and the version you use?

Kyungpyo-Kim commented 4 years ago

This is my environment information. I hope it can help you.

My PC Info

CUDA 10.0 Installation

sudo cuda_10.0.130_410.48_linux.run
$ export PATH=/usr/local/cuda-10.0/bin:$PATH
$ export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64:$LD_LIBRARY_PATH

cuDNN 7.4.2 Installation

sudo dpkg -i ./libcudnn7-dev_7.4.2.24-1+cuda10.0_amd64.deb 
sudo dpkg -i ./libcudnn7-doc_7.4.2.24-1+cuda10.0_amd64.deb 
sudo dpkg -i ./libcudnn7_7.4.2.24-1+cuda10.0_amd64.deb

cupti

sudo apt install libcupti-dev -y

Tensorflow-gpu 1.14

pip3 install tensorflow-gpu opencv-python jupyter matplotlib tqdm scipy googledrivedownloader
pip3 uninstall numpy
pip3 install --upgrade numpy==1.16.1

Demo results:

Test inference speed icnet_cityscapes_trainval_90k: Average time: 0.5310, about 1.883186 fps icnet_cityscapes_train_30k: Average time: 0.5302, about 1.885943 fps