Skip to main content
Help us improve Python packaging – donate today!

Fully Convolutional Networks

Project Description

fcn - Fully Convolutional Networks
==================================

[![](https://badge.fury.io/gh/wkentaro%2Ffcn.svg)](https://badge.fury.io/gh/wkentaro%2Ffcn)
[![](https://api.travis-ci.org/wkentaro/fcn.svg?branch=master)](https://travis-ci.org/wkentaro/fcn)

This is [Chainer](https://github.com/pfnet/chainer.git) implementation of
[fcn.berkeley.vision.org](https://github.com/shelhamer/fcn.berkeleyvision.org.git).


Installation
------------

```bash
pip install --upgrade setuptools

pip install fcn
```


Inference
---------

Inference is done as below:

```bash
# Download sample image
wget https://farm2.staticflickr.com/1522/26471792680_a485afb024_z_d.jpg -O sample.jpg

# forwaring of the networks
fcn_infer.py --img-files sample.jpg --gpu -1 # cpu mode
fcn_infer.py --img-files sample.jpg # gpu mode
```

<img src="static/fcn8s_26471792680.jpg" width="80%" >

Original Image: <https://www.flickr.com/photos/faceme/26471792680/>


Training
--------

```bash
cd examples/voc
./download_dataset.py

./train_fcn32s.py --gpu 0
# ./train_fcn16s.py --gpu 0
# ./train_fcn8s.py --gpu 0
# ./train_fcn8s_atonce.py --gpu 0
```

The accuracy of original implementation is computed with (`evaluate.py`) after converting the caffe model to chainer one
using `convert_caffe_to_chainermodel.py`.\
You can download vgg16 model from here: [`vgg16_from_caffe.npz`](https://drive.google.com/open?id=0B9P1L--7Wd2vRy1XYnRSa1hNSW8).

**FCN32s**

| Implementation | Accuracy | Accuracy Class | Mean IU | FWAVACC | Model File |
|:--------------:|:--------:|:--------------:|:-------:|:-------:|:----------:|
| [Original](https://github.com/shelhamer/fcn.berkeleyvision.org/tree/master/voc-fcn32s) | 90.4810 | 76.4824 | 63.6261 | 83.4580 | [`fcn32s_from_caffe.npz`](https://drive.google.com/uc?id=0B9P1L--7Wd2vTElpa1p3WFNDczQ) |
| Ours (using `vgg16_from_caffe.npz`) | **90.5668** | **76.8740** | **63.8180** | **83.5067** | [`fcn32s_voc_iter00092000.npz`](https://drive.google.com/uc?0B9P1L--7Wd2vRTQzQl8xcUI5Uk0) |

**FCN16s**

| Implementation | Accuracy | Accuracy Class | Mean IU | FWAVACC | Model File |
|:--------------:|:--------:|:--------------:|:-------:|:-------:|:----------:|
| [Original](https://github.com/shelhamer/fcn.berkeleyvision.org/tree/master/voc-fcn16s) | 90.9971 | **78.0710** | 65.0050 | 84.2614 | [`fcn16s_from_caffe.npz`](https://drive.google.com/uc?id=0B9P1L--7Wd2vcnBiXzZTcG9FU3c) |
| Ours (using `fcn32s_from_caffe.npz`) | 90.9671 | 78.0617 | 65.0911 | 84.2604 | [`fcn16s_voc_using_fcn32s_from_caffe_iter00032000.npz`](https://drive.google.com/uc?id=0B9P1L--7Wd2vNTFyZDlXel9ZZms) |
| Ours (using `fcn32s_voc_iter00092000.npz`) | **91.1009** | 77.2522 | **65.3628** | **84.3675** | [`fcn16s_voc_iter00100000.npz`](https://drive.google.com/uc?id=0B9P1L--7Wd2vZ1ZUYTJhRkZ1WTg) |

**FCN8s**

| Implementation | Accuracy | Accuracy Class | Mean IU | FWAVACC | Model File |
|:--------------:|:--------:|:--------------:|:-------:|:-------:|:----------:|
| [Original](https://github.com/shelhamer/fcn.berkeleyvision.org/tree/master/voc-fcn8s) | 91.2212 | 77.6146 | 65.5126 | 84.5445 | [`fcn8s_from_caffe.npz`](https://drive.google.com/uc?id=0B9P1L--7Wd2vb0cxV0VhcG1Lb28) |
| Ours (using `fcn16s_from_caffe.npz`) | 91.2513 | 77.1490 | 65.4789 | 84.5460 | [`fcn8s_voc_using_fcn16s_from_caffe_iter00016000.npz`](https://drive.google.com/uc?id=0B9P1L--7Wd2vdVpRN253el9fdzA) |
| Ours (using `fcn16s_voc_iter00100000.npz`) | **91.2608** | **78.1484** | **65.8444** | **84.6447** | [`fcn8s_voc_iter00072000.npz`](https://drive.google.com/uc?id=0B9P1L--7Wd2vWG5MeUEwWmxudU0) |

**FCN8sAtOnce**

| Implementation | Accuracy | Accuracy Class | Mean IU | FWAVACC | Model File |
|:--------------:|:--------:|:--------------:|:-------:|:-------:|:----------:|
| [Original](https://github.com/shelhamer/fcn.berkeleyvision.org/tree/master/voc-fcn8s-atonce) | **91.1288** | **78.4979** | **65.3998** | **84.4326** | [`fcn8s-atonce_from_caffe.npz`](https://drive.google.com/uc?id=0B9P1L--7Wd2vZ1RJdXotZkNhSEk) |
| Ours (using `vgg16_from_caffe.npz`) | 91.0883 | 77.3528 | 65.3433 | 84.4276 | [`fcn8s-atonce_voc_iter00056000.npz`](https://drive.google.com/uc?id=0B9P1L--7Wd2vcl9STGhJY1J4WUE) |

<img src="examples/voc/static/fcn32s_iter00092000.jpg" width="30%" /> <img src="examples/voc/static/fcn16s_iter00100000.jpg" width="30%" /> <img src="examples/voc/static/fcn8s_iter00072000.jpg" width="30%" />

Left to right, **FCN32s**, **FCN16s** and **FCN8s**, which are fully trained using this repo. See above tables to see the accuracy.

Release history Release notifications

History Node

6.3.0

History Node

6.2.0

History Node

6.1.10

History Node

6.1.9

History Node

6.1.8

History Node

6.1.7

History Node

6.1.6

History Node

6.1.5

History Node

6.1.4

History Node

6.1.3

History Node

6.1.2

History Node

6.1.1

History Node

6.1.0

History Node

6.0.6

History Node

6.0.5

This version
History Node

6.0.4

History Node

6.0.3

History Node

6.0.2

History Node

6.0.1

History Node

6.0.0

History Node

5.8.1

History Node

5.8.0

History Node

5.7.1

History Node

5.7.0

History Node

5.6.0

History Node

5.5.0

History Node

5.4.0

History Node

5.3.0

History Node

5.2.1

History Node

5.2.0

History Node

5.1.0

History Node

5.0.2

History Node

5.0.1

History Node

5.0.0

History Node

4.0.2

History Node

4.0.1

History Node

4.0.0

History Node

3.2.0

History Node

3.1.0

History Node

3.0.0

History Node

2.0.0

History Node

1.3.5

History Node

1.3.4

History Node

1.3.3

History Node

1.3.2

History Node

1.3.1

History Node

1.3.0

History Node

1.2.9

History Node

1.2.8

History Node

1.2.7

History Node

1.2.6

History Node

1.2.5

History Node

1.2.4

History Node

1.2.3

History Node

1.2.2

History Node

1.2.1

History Node

1.2.0

History Node

1.1.3

History Node

1.1.2

History Node

1.1.1

History Node

1.1.0

History Node

1.0.1

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, size & hash SHA256 hash help File type Python version Upload date
fcn-6.0.4.tar.gz (1.2 MB) Copy SHA256 hash SHA256 Source None Jul 17, 2017

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging CloudAMQP CloudAMQP RabbitMQ AWS AWS Cloud computing Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page