A VGG16 inference module comparison between
- tensorflow
- numpy
- pycuda
- required dependencies were installed with miniforge using
conda install pillow pycuda tensorflow-gpu
- VGG16 weights pretrained on ImageNet were downloaded from here and placed in the same directory as the code in this repository
$ ./vgg.py tf /path/to/image_to_classify # run inference using tensorflow
$ ./vgg.py numpy /path/to/image_to_classify # run inference using numpy implementation
$ ./vgg.py pycuda /path/to/image_to_classify # run inference using pycuda implementation