peaknet

A peaknet API with pytorch backbone

conda environment recipe

conda activate ana-1.4.22

To build/test the python, cd to the directory containing setup.py and execute the following commands:

• mkdir -p install/lib/python2.7/site-packages
• export PYTHONPATH=`pwd`/install/lib/python2.7/site-packages
• python setup.py develop --prefix=`pwd`/install

Example

from peaknet.Peaknet import Peaknet
peaknet = Peaknet(use_cuda=True) # Init a Peaknet instance
peaknet.loadDefaultCFG() # Load newpeaksv10 network and pretrained weights 

Default model file lives at /reg/common/package/peaknet/model.pt

API

predict

peaknet.predict( imgs )

imgs is a numpy array with dimensions (n,m,h,w). imgs will be treated as a stack of nxm tiles.

train (for client)

peaknet.train( imgs, labels, box_size = 7 )

imgs is a numpy array with dimensions (n,m,h,w). imgs will be treated as a stack of nxm tiles. labels is a list of tutple of length n. Each item in the list is a tutple of three numpy arrays s, r, c, where s is an array of integers 0~(m-1).

model access

peaknet.model

returns the current model

update model

peaknet.updateModel( newModel )

replaces the current model with newModel, including the network and the weights

update grad

peaknet.updateGrad( newModel )

replaces gradients in the current model with that from newModel. newModel must have same network as current model.

optimize

peaknet.optimize()

performs one step of SGD optimization.