Cnn-Hypercolumn
IN coming:
Inception V3
main code: Cnn-Hypercolumn.py
more information in Cnn-ypercolumn
Content
model and usage demo: see vgg-16_keras.py
weights: vgg16_weights.h5
model and usage demo: see vgg-19_keras.py
weights: vgg19_weights.h5
Instruction
to get started, you will need to proceed this way:
pip install -r requirements.txt # Install dependencies
if you want to work in a virtual environement temporarly:
sudo pip install virtualenv # This may already be installed
virtualenv .env # Create a virtual environment
source .env/bin/activate # Activate the virtual environment
pip install -r requirements.txt # Install dependencieswork for a while....
deactivate
WARNING
The last part in main.py should be run on a computer with enough RAM or python will be killed
# Simple hypercolumn pixel clustering
m = hc.transpose(1,2,0).reshape(50176, -1)
kmeans = cluster.KMeans(n_clusters=2, max_iter=300, n_jobs=5, precompute_distances=True)
cluster_labels = kmeans .fit_predict(m)
imcluster = np.zeros((224,224))
imcluster = imcluster.reshape((224*224,))
imcluster = cluster_labels
plt.imshow(imcluster.reshape(224, 224), cmap="hot")
#plt.show()OpenCv - cv2
direct way
To install and use cv2, please follow the instruction at the following address
http://www.pyimagesearch.com/2015/06/15/install-opencv-3-0-and-python-2-7-on-osx/
Anaconda
conda install opencv