Metadata-Version: 2.1
Name: torchray
Version: 1.0.0.1
Summary: TorchRay is a PyTorch library of visualization methods for convnets.
Home-page: https://github.com/facebookresearch/TorchRay
Author: Andrea Vedaldi
Author-email: vedaldi@fb.com
License: Creative Commons Attribution-Noncommercial 4.0 International
Download-URL: https://github.com/ruthcfong/TorchRay/archive/v1.0.0.1.tar.gz
Description: # TorchRay
        
        The *TorchRay* package implements several visualization methods for deep
        convolutional neural networks using PyTorch. In this release, TorchRay focuses
        on *attribution*, namely the problem of determining which part of the input,
        usually an image, is responsible for the value computed by a neural network.
        
        *TorchRay* is research oriented: in addition to implementing well known
        techniques form the literature, it provides code for reproducing results that
        appear in several papers, in order to support *reproducible research*.
        
        *TorchRay* was initially developed to support the paper:
        
        * *Understanding deep networks via extremal perturbations and smooth masks.*
          Fong, Patrick, Vedaldi.
          Proceedings of the International Conference on Computer Vision (ICCV), 2019.
        
        ### Captum vs. TorchRay
        Facebook also released *[Captum](https://github.com/pytorch/captum)*, another interpretability library built on PyTorch. *Captum* provides broader support for other domains beyond vision and provides out-of-the-box interpretability methods for any model. While we also provide out-of-the-box methods, *TorchRay* primarily focuses on making interpretability research reproducible by providing a set of standard models and benchmarks, focusing on computer vision.
        
        ## Examples
        
        The package contains several usage examples in the
        [`examples`](https://github.com/facebookresearch/TorchRay/tree/master/examples)
        subdirectory.
        
        Here is a complete example for using GradCAM:
        
        ```python
        from torchray.attribution.grad_cam import grad_cam
        from torchray.benchmark import get_example_data, plot_example
        
        # Obtain example data.
        model, x, category_id, _ = get_example_data()
        
        # Grad-CAM backprop.
        saliency = grad_cam(model, x, category_id, saliency_layer='features.29')
        
        # Plots.
        plot_example(x, saliency, 'grad-cam backprop', category_id)
        ```
        
        ## Requirements
        
        TorchRay requires:
        
        * Python 3.4 or greater
        * pytorch 1.1.0 or greater
        * matplotlib
        
        For benchmarking, it also requires:
        
        * torchvision 0.3.0 or greater
        * pycocotools
        * mongodb (suggested)
        * pymongod (suggested)
        
        On Linux/macOS, using conda you can install
        
        ```bash
        while read requirement; do conda install \
        -c defaults -c pytorch -c conda-forge --yes $requirement; done <<EOF
        pytorch>=1.1.0
        pycocotools
        torchvision>=0.3.0
        mongodb
        pymongo
        EOF
        ```
        
        ## Installing TorchRay
        
        Using `pip`:
        
        ```shell
        pip install torchray
        ```
        
        From source:
        
        ```shell
        python setup.py install
        ```
        
        or
        
        ```shell
        pip install .
        ```
        
        ## Full documentation
        
        The full documentation can be found
        [here](https://facebookresearch.github.io/TorchRay).
        
        ## Changes
        
        See the [CHANGELOG](CHANGELOG.md).
        
        ## Join the TorchRay community
        
        See the [CONTRIBUTING](CONTRIBUTING.md) file for how to help out.
        
        ## The team
        
        TorchRay has been primarily developed by Ruth C. Fong and Andrea Vedaldi.
        
        ## Acknowledgements
        
        We'd like to acknowledge Jianming Zhang and Samuel Albanie for help on re-implementing the Pointing Game and Chris Olah and the OpenAI Clarity team for helpful discussions.
        
        ## License
        
        TorchRay is CC-BY-NC licensed, as found in the [LICENSE](LICENSE) file.
        
        ## Reference
        
        If you use our library, please cite the following:
        ```
        @InProceedings{fong_iccv_2019,
        author = {Fong, Ruth and Patrick, Mandela and Vedaldi, Andrea},
        title = {Understanding Deep Networks via Extremal Perturbations and Smooth Masks},
        booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
        year = {2019}
        }
        ```
        
        
Platform: UNKNOWN
Classifier: Development Status :: 5 - Production/Stable
Classifier: Intended Audience :: Developers
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.4
Classifier: Programming Language :: Python :: 3.5
Classifier: Programming Language :: Python :: 3.6
Classifier: Programming Language :: Python :: 3.7
Classifier: Topic :: Scientific/Engineering
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: Software Development
Classifier: Topic :: Software Development :: Libraries
Classifier: Topic :: Software Development :: Libraries :: Python Modules
Requires-Python: >=3.4
Description-Content-Type: text/markdown
