def run(self, ips, imgs, para=None, preview=False):
if len(ips.imgs) == 1: ips.img[:] = ips.snap
key = {'chans': None, 'grade': para['grade'], 'w': para['w']}
key['items'] = [i for i in ['ori', 'blr', 'sob', 'eig'] if para[i]]
slir, slic = ips.rect
labs = [i[slir, slic] for i in imgs]
ori = ips.back.imgs
oris = [i[slir, slic] for i in ori]
self.app.info('extract features...')
feat, lab, key = feature.get_feature(oris,
labs,
key,
callback=self.progress)
self.app.info('training data...')
# self.progress(None, 1)
model = self.classify(para)
model.fit(feat, lab)
self.app.info('predict data...')
if preview:
return feature.get_predict(oris,
model,
key,
labs,
callback=self.progress)
if len(imgs) == 1: ips.swap()
outs = feature.get_predict(oris, model, key, callback=self.progress)
nips = Image(outs, ips.title + 'rst')
nips.range, nips.lut = ips.range, ips.lut
nips.back, nips.mode = ips.back, 0.4
self.app.show_img(nips)
global model_para
model_para = model, key
def run(self, ips, imgs, para=None):
if '/' in para['predictor']: path = para['predictor']
else: path = self.root + '/' + para['predictor']
model, key = joblib.load(path)
if not para['slice']: imgs = [ips.img]
slir, slic = ips.rect
imgs = [i[slir, slic] for i in imgs]
rst = feature.get_predict(imgs, model, key, callback=self.progress)
if rst is None:
return self.app.alert('image channels dismatch this predictor!')
ips = Image(rst, ips.title + '-mark')
ips.range = ips.get_updown('all', 'one', step=512)
self.app.show_img(ips)
def run(self, ips, imgs, para = None):
if not para['slice']: imgs = [ips.img]
labels = []
for i in range(len(imgs)):
self.progress(i, len(imgs))
con = 1 if para['con']=='4-Connect' else 2
idx = connect.connect_graph(imgs[i], con, para['back'])
idx = connect.mapidx(idx)
cmap = render.node_render(idx, para['colors'], 10)
lut = np.ones(imgs[i].max()+1, dtype=np.uint8)
lut[0] = 0
for j in cmap: lut[j] = cmap[j]
labels.append(lut[imgs[i]])
ips = Image(labels, ips.title+'-render')
ips.range = (0, para['colors'])
self.app.show_img(ips)