This is Chainer implementation of fcn.berkeleyvision.org.

• Provide FCN8s model for Chainer. [v1.0.0]
• Copy caffemodel to chainermodel. [v1.0.0]
• Forwarding with Chainer for pascal dataset. [v1.0.0]
• Training with Chainer for pascal dataset. [v2.0.0]
• Training for APC2015 dataset. [v3.0.0]

pip install fcn
fcn_install_trained_data.py

Forwarding is done as below, and computation graph is here.

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

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

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

git clone https://github.com/wkentaro/fcn.git
cd fcn

python setup.py install
fcn_install_trained_data.py

You need to download pascal VOC2012 dataset from here, and install it as below construction:

- fcn - data - pascal - VOC2012 -- JPEGImages
                                 - SegmentationClass
                                 - ...

Forwarding with trained model in caffe is done as below:

# This downloads caffemodel and convert it to chainermodel
./scripts/caffe_to_chainermodel.py

# forwarding of the networks
./scripts/fcn_forward.py --img-files data/pascal/VOC2012/JPEGImages/2007_000129.jpg

Original Image: http://host.robots.ox.ac.uk/pascal/VOC/voc2012/

./scripts/fcn_train.py

Currently we support only training FCN32s. The learning curve looks like below:

Forwarding with fcn32s_60000.chainermodel ends with below result: