Minimalistic python package to extract deep learning features from a wide variety of pretrained models in Pytorch.
Dependencies:
- pytorch (v >= 1.0.0)
- torchvision
To install use pip:
$ pip install imfeatures
Or clone the repo:
$ git clone https://github.com/resbyte/imfeatures.git
$ python setup.py install
- Imports
import imfeatures
import torch
- create feature extractor, here
resnet50
, with pretrained weights
feature_extractor = imfeatures.Features('resnet50',pretrained=True)
- random image of size
224x224x3
x = torch.randn([1,3,224,224])
- features
features = feature_extractor(x)
print(features.shape)
Output features will be of shape : [1, 2048, 1, 1]
Abhinav Dadhich