[PYTHON] I tried object detection with YOLO v3 (TensorFlow 2.1) on the GPU of windows!

environment

windows7 64bit Gefore GTX 680MX GPU anaconda

1. 1. CUDA installation

Download the CUDA Toolkit 10.1 update 2 version from the following site (TensorFlow 2.1.0 is supported because of CUDA Toolkit 10.1) https://developer.nvidia.com/cuda-toolkit-archive

Do not select Visual Studio Intergration for installation

If the installation is successful, execute nvcc -V and the following result will be displayed.

C:\Windows\System32>nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Sun_Jul_28_19:12:52_Pacific_Daylight_Time_2019
Cuda compilation tools, release 10.1, V10.1.243

2. cuDNN installation

Select and download cuDNN for CUDA 10.1 from the following site (Requires NVIDIA account) https://developer.nvidia.com/cudnn

Unzip the zip file and Rename the cuda folder to cuda765 and Copy to the path "C: \ Program Files \ NVIDIA GPU Computing Toolkit \ CUDA \ v10.1"

After copying

3. Environment variable settings

Add the following cuDNN path to PATH

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\cuda765\bin

4. Create keras_work with conda

conda create -n keras_work
activate keras_work

5. Install tensoflow

conda install tensorflow

6. Confirmation

(base) C:\Users\mac>activate keras_work

(keras_work) C:\Users\mac>python
Python 3.7.7 (default, Apr 15 2020, 05:09:04) [MSC v.1916 64 bit (AMD64)] :: Ana
conda, Inc. on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
2020-05-06 19:46:04.480481: I tensorflow/stream_executor/platform/default/dso_lo
ader.cc:44] Successfully opened dynamic library cudart64_101.dll
>>> print(tf.__version__)
2.1.0
>>> print(tf.test.is_gpu_available())
・ ・ ・
physical GPU (device: 0, name: GeForce GTX 680MX, pci bus id: 0000:01:00.0, compute capability: 3.0)
True

7. YOLO V3 material preparation

Get source from Github

cd c:\temp
git clone https://github.com/zzh8829/yolov3-tf2.git
cd yolov3-tf2

Download yolo3.weight from pjreddie.com

• It takes 2 hours

wget https://pjreddie.com/media/files/yolov3.weights --no-check-certificate

Since it is slow to get yolov3.weights from the above pjreddie.com, you can also download the following URL. https://pan.baidu.com/s/1G2Qh-V8kyLOq4oDbTwK6HQ Proposal (password): vogw The file is "yolo_tf2.1 \ data \ yolov3.weights"

Move yolo3.weight file to yolov3-tf2 path Confirm that the file has been moved

(keras_work) C:\temp\yolov3-tf2>dir /B *.weights
yolov3.weights

8. Convert (convert pre-trained darknet weight)

python convert.py --weights ./yolov3.weights --output ./checkpoints/yolov3.tf

9. detection

python detect.py --image ./data/street.jpg