Image-Py / planer-zoo

toolbox of planer

planer-store

Planer is a light-weight CNN framework implemented in pure Numpy-like interface. It can run only with Numpy. Or change different backends. (Cupy accelerated with CUDA, ClPy accelerated with OpenCL).

planer-store is a toolbox based on planer. Supportting many models, we can use them easily:

import planer_store as plas
model = plas.load_model('xxx')

model.xxx ...

crnn-ocr

import planer_store as plas
model = plas.load_model('crnn_ocr')

img = imread('planer.jpg')
text = model.recognize(img)

# show img and result

face-68-key

import planer_store as plas
model = plas.load_model('face_68_key')

face = imread('face.jpg')
rc = model.get_face_key(face)
thin = model.face_adjust(face, fac=-10)
fat = model.face_adjust(face, fac=10)

# show face, thin, fat

hed-edge

import planer_store as plas
model = plas.load_model('hed_edge')

face = imread('edge.jpg')
edge = model.hed_edge(img)

# show face, edge

object-recognize

import planer_store as plas
model = plas.load_model('resnet18')

img = imread('bus.jpg')
obj = model.recognize(img)

# show img and result

high-resolution

import planer_store as plas
model = plas.load_model('ESRGAN')

img = imread('girl.jpg')
high = model.super_resolution(img)

# show img and high resolution result

cellpose

import planer_store as plas
model = plas.load_model('cellpose')

img = 255 - imread('cell.png')[:,:,0]
flow = count_flow(img)
lab = flow2msk(flow, level=0.2)
edge = draw_edge(img, lab)
rgb = rgb_mask(img, lab)

# show img, flow, edge, rgbmsk

Contribution

welcom to contribute new models.

1. use torch to train, and export as onnx
2. planer.onnx2planer('xxx.onnx')