Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/shrubb/box-convolutions
PyTorch code for the "Deep Neural Networks with Box Convolutions" paper
https://github.com/shrubb/box-convolutions
Last synced: about 2 months ago
JSON representation
PyTorch code for the "Deep Neural Networks with Box Convolutions" paper
- Host: GitHub
- URL: https://github.com/shrubb/box-convolutions
- Owner: shrubb
- License: apache-2.0
- Created: 2018-10-26T09:21:09.000Z (almost 6 years ago)
- Default Branch: master
- Last Pushed: 2020-01-20T06:52:23.000Z (over 4 years ago)
- Last Synced: 2024-07-11T13:58:18.207Z (2 months ago)
- Language: C++
- Homepage:
- Size: 104 KB
- Stars: 509
- Watchers: 17
- Forks: 35
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
Box Convolution Layer for ConvNets
==================================
Single-box-conv network (from `examples/mnist.py`) learns patterns on MNIST
# What This Is
This is a PyTorch implementation of the box convolution layer as introduced in the 2018 NeurIPS [paper](https://papers.nips.cc/paper/7859-deep-neural-networks-with-box-convolutions):
Burkov, E., & Lempitsky, V. (2018) **Deep Neural Networks with Box Convolutions**. *Advances in Neural Information Processing Systems 31*, 6214-6224.
# How to Use
## Installing
```bash
python3 -m pip install git+https://github.com/shrubb/box-convolutions.git
python3 -m box_convolution.test # if throws errors, please open a GitHub issue
```
To uninstall:
```bash
python3 -m pip uninstall box_convolution
```
Tested on Ubuntu 18.04.2, Python 3.6, PyTorch 1.0.0, GCC {4.9, 5.5, 6.5, 7.3}, CUDA 9.2. Other versions (e.g. macOS or Python 2.7 or CUDA 8 or CUDA 10) should work too, but I haven't checked. If something doesn't build, please open a Github issue.
Known issues (see [this chat](https://github.com/shrubb/box-convolutions/issues/2)):
* CUDA 9/9.1 + GCC 6 isn't supported due to a bug in NVCC.
You can specify a different compiler with `CC` environment variable:
```bash
CC=g++-7 python3 -m pip install git+https://github.com/shrubb/box-convolutions.git
```
## Using
```python3
import torch
from box_convolution import BoxConv2d
box_conv = BoxConv2d(16, 8, 240, 320)
help(BoxConv2d)
```
Also, there are usage examples in `examples/`.
# Quick Tour of Box convolutions
You may want to see our [poster](https://yadi.sk/i/LNnMrj6FwbOc9A).
### Why reinvent the old convolution?
`3×3` convolutions are too small ⮕ receptive field grows too slow ⮕ ConvNets have to be very deep.
This is especially undesirable in dense prediction tasks (*segmentation, depth estimation, object detection, ...*).
Today people solve this by
* dilated/deformable convolutions (can bring artifacts or degrade to `1×1` conv; almost always filter high-frequency);
* "global" spatial pooling layers (usually too constrained, fixed size, not "fully convolutional").
### How does it work?
Box convolution layer is a basic *depthwise convolution* (i.e. `Conv2d` with `groups=in_channels`) but with special kernels called *box kernels*.
A box kernel is a rectangular averaging filter. That is, filter values are fixed and unit! Instead, we learn four parameters per rectangle − its size and offset:
### Any success stories?
One example: there is an efficient semantic segmentation model [**ENet**](https://github.com/e-lab/ENet-training). It's a classical hourglass architecture stacked of dozens ResNet-like blocks (left image).
Let's replace some of these blocks by our "box convolution block" (right image).
First we replaced every second block with a box convolution block (*Box*ENet in the paper). The model became
* more accurate,
* faster,
* lighter
* **without dilated convolutions**.
Then, we replaced **every** residual block (except the down- and up-sampling ones)! The result, *BoxOnly*ENet, is
* a **ConvNet almost without** (traditional learnable weight) **convolutions**,
* **2** times less operations,
* **3** times less parameters,
* still **more accurate** than ENet!