import sys
import os
import numpy as np
import torch
import matplotlib
matplotlib.use('agg')
import matplotlib.pyplot as plt
def getFreeId():
import pynvml
pynvml.nvmlInit()
def getFreeRatio(id):
handle = pynvml.nvmlDeviceGetHandleByIndex(id)
use = pynvml.nvmlDeviceGetUtilizationRates(handle)
ratio = 0.5*(float(use.gpu+float(use.memory)))
return ratio
deviceCount = pynvml.nvmlDeviceGetCount()
available = []
for i in range(deviceCount):
if getFreeRatio(i)<70:
available.append(i)
gpus = ''
for g in available:
gpus = gpus+str(g)+','
gpus = gpus[:-1]
return gpus
def setgpu(gpuinput):
freeids = getFreeId()
if gpuinput=='all':
gpus = freeids
else:
gpus = gpuinput
if any([g not in freeids for g in gpus.split(',')]):
raise ValueError('gpu'+g+'is being used')
print('using gpu '+gpus)
os.environ['CUDA_VISIBLE_DEVICES']=gpus
return len(gpus.split(','))
class Logger(object):
def __init__(self,logfile):
self.terminal = sys.stdout
self.log = open(logfile, "a")
def write(self, message):
self.terminal.write(message)
self.log.write(message)
def flush(self):
#this flush method is needed for python 3 compatibility.
#this handles the flush command by doing nothing.
#you might want to specify some extra behavior here.
pass
def split4(data, max_stride, margin):
splits = []
data = torch.Tensor.numpy(data)
_,c, z, h, w = data.shape
w_width = np.ceil(float(w / 2 + margin)/max_stride).astype('int')*max_stride
h_width = np.ceil(float(h / 2 + margin)/max_stride).astype('int')*max_stride
pad = int(np.ceil(float(z)/max_stride)*max_stride)-z
leftpad = pad/2
pad = [[0,0],[0,0],[leftpad,pad-leftpad],[0,0],[0,0]]
data = np.pad(data,pad,'constant',constant_values=-1)
data = torch.from_numpy(data)
splits.append(data[:, :, :, :h_width, :w_width])
splits.append(data[:, :, :, :h_width, -w_width:])
splits.append(data[:, :, :, -h_width:, :w_width])
splits.append(data[:, :, :, -h_width:, -w_width:])
return torch.cat(splits, 0)
def combine4(output, h, w):
splits = []
for i in range(len(output)):
splits.append(output[i])
output = np.zeros((
splits[0].shape[0],
h,
w,
splits[0].shape[3],
splits[0].shape[4]), np.float32)
h0 = output.shape[1] / 2
h1 = output.shape[1] - h0
w0 = output.shape[2] / 2
w1 = output.shape[2] - w0
splits[0] = splits[0][:, :h0, :w0, :, :]
output[:, :h0, :w0, :, :] = splits[0]
splits[1] = splits[1][:, :h0, -w1:, :, :]
output[:, :h0, -w1:, :, :] = splits[1]
splits[2] = splits[2][:, -h1:, :w0, :, :]
output[:, -h1:, :w0, :, :] = splits[2]
splits[3] = splits[3][:, -h1:, -w1:, :, :]
output[:, -h1:, -w1:, :, :] = splits[3]
return output
def split8(data, max_stride, margin):
splits = []
if isinstance(data, np.ndarray):
c, z, h, w = data.shape
else:
_,c, z, h, w = data.size()
z_width = np.ceil(float(z / 2 + margin)/max_stride).astype('int')*max_stride
w_width = np.ceil(float(w / 2 + margin)/max_stride).astype('int')*max_stride
h_width = np.ceil(float(h / 2 + margin)/max_stride).astype('int')*max_stride
for zz in [[0,z_width],[-z_width,None]]:
for hh in [[0,h_width],[-h_width,None]]:
for ww in [[0,w_width],[-w_width,None]]:
if isinstance(data, np.ndarray):
splits.append(data[np.newaxis, :, zz[0]:zz[1], hh[0]:hh[1], ww[0]:ww[1]])
else:
splits.append(data[:, :, zz[0]:zz[1], hh[0]:hh[1], ww[0]:ww[1]])
if isinstance(data, np.ndarray):
return np.concatenate(splits, 0)
else:
return torch.cat(splits, 0)
def combine8(output, z, h, w):
splits = []
for i in range(len(output)):
splits.append(output[i])
output = np.zeros((
z,
h,
w,
splits[0].shape[3],
splits[0].shape[4]), np.float32)
z_width = z / 2
h_width = h / 2
w_width = w / 2
i = 0
for zz in [[0,z_width],[z_width-z,None]]:
for hh in [[0,h_width],[h_width-h,None]]:
for ww in [[0,w_width],[w_width-w,None]]:
output[zz[0]:zz[1], hh[0]:hh[1], ww[0]:ww[1], :, :] = splits[i][zz[0]:zz[1], hh[0]:hh[1], ww[0]:ww[1], :, :]
i = i+1
return output
def split16(data, max_stride, margin):
splits = []
_,c, z, h, w = data.size()
z_width = np.ceil(float(z / 4 + margin)/max_stride).astype('int')*max_stride
z_pos = [z*3/8-z_width/2,
z*5/8-z_width/2]
h_width = np.ceil(float(h / 2 + margin)/max_stride).astype('int')*max_stride
w_width = np.ceil(float(w / 2 + margin)/max_stride).astype('int')*max_stride
for zz in [[0,z_width],[z_pos[0],z_pos[0]+z_width],[z_pos[1],z_pos[1]+z_width],[-z_width,None]]:
for hh in [[0,h_width],[-h_width,None]]:
for ww in [[0,w_width],[-w_width,None]]:
splits.append(data[:, :, zz[0]:zz[1], hh[0]:hh[1], ww[0]:ww[1]])
return torch.cat(splits, 0)
def combine16(output, z, h, w):
splits = []
for i in range(len(output)):
splits.append(output[i])
output = np.zeros((
z,
h,
w,
splits[0].shape[3],
splits[0].shape[4]), np.float32)
z_width = z / 4
h_width = h / 2
w_width = w / 2
splitzstart = splits[0].shape[0]/2-z_width/2
z_pos = [z*3/8-z_width/2,
z*5/8-z_width/2]
i = 0
for zz,zz2 in zip([[0,z_width],[z_width,z_width*2],[z_width*2,z_width*3],[z_width*3-z,None]],
[[0,z_width],[splitzstart,z_width+splitzstart],[splitzstart,z_width+splitzstart],[z_width*3-z,None]]):
for hh in [[0,h_width],[h_width-h,None]]:
for ww in [[0,w_width],[w_width-w,None]]:
output[zz[0]:zz[1], hh[0]:hh[1], ww[0]:ww[1], :, :] = splits[i][zz2[0]:zz2[1], hh[0]:hh[1], ww[0]:ww[1], :, :]
i = i+1
return output
def split32(data, max_stride, margin):
splits = []
_,c, z, h, w = data.size()
z_width = np.ceil(float(z / 2 + margin)/max_stride).astype('int')*max_stride
w_width = np.ceil(float(w / 4 + margin)/max_stride).astype('int')*max_stride
h_width = np.ceil(float(h / 4 + margin)/max_stride).astype('int')*max_stride
w_pos = [w*3/8-w_width/2,
w*5/8-w_width/2]
h_pos = [h*3/8-h_width/2,
h*5/8-h_width/2]
for zz in [[0,z_width],[-z_width,None]]:
for hh in [[0,h_width],[h_pos[0],h_pos[0]+h_width],[h_pos[1],h_pos[1]+h_width],[-h_width,None]]:
for ww in [[0,w_width],[w_pos[0],w_pos[0]+w_width],[w_pos[1],w_pos[1]+w_width],[-w_width,None]]:
splits.append(data[:, :, zz[0]:zz[1], hh[0]:hh[1], ww[0]:ww[1]])
return torch.cat(splits, 0)
def combine32(splits, z, h, w):
output = np.zeros((
z,
h,
w,
splits[0].shape[3],
splits[0].shape[4]), np.float32)
z_width = int(np.ceil(float(z) / 2))
h_width = int(np.ceil(float(h) / 4))
w_width = int(np.ceil(float(w) / 4))
splithstart = splits[0].shape[1]/2-h_width/2
splitwstart = splits[0].shape[2]/2-w_width/2
i = 0
for zz in [[0,z_width],[z_width-z,None]]:
for hh,hh2 in zip([[0,h_width],[h_width,h_width*2],[h_width*2,h_width*3],[h_width*3-h,None]],
[[0,h_width],[splithstart,h_width+splithstart],[splithstart,h_width+splithstart],[h_width*3-h,None]]):
for ww,ww2 in zip([[0,w_width],[w_width,w_width*2],[w_width*2,w_width*3],[w_width*3-w,None]],
[[0,w_width],[splitwstart,w_width+splitwstart],[splitwstart,w_width+splitwstart],[w_width*3-w,None]]):
output[zz[0]:zz[1], hh[0]:hh[1], ww[0]:ww[1], :, :] = splits[i][zz[0]:zz[1], hh2[0]:hh2[1], ww2[0]:ww2[1], :, :]
i = i+1
return output
def split64(data, max_stride, margin):
splits = []
_,c, z, h, w = data.size()
z_width = np.ceil(float(z / 4 + margin)/max_stride).astype('int')*max_stride
w_width = np.ceil(float(w / 4 + margin)/max_stride).astype('int')*max_stride
h_width = np.ceil(float(h / 4 + margin)/max_stride).astype('int')*max_stride
z_pos = [z*3/8-z_width/2,
z*5/8-z_width/2]
w_pos = [w*3/8-w_width/2,
w*5/8-w_width/2]
h_pos = [h*3/8-h_width/2,
h*5/8-h_width/2]
for zz in [[0,z_width],[z_pos[0],z_pos[0]+z_width],[z_pos[1],z_pos[1]+z_width],[-z_width,None]]:
for hh in [[0,h_width],[h_pos[0],h_pos[0]+h_width],[h_pos[1],h_pos[1]+h_width],[-h_width,None]]:
for ww in [[0,w_width],[w_pos[0],w_pos[0]+w_width],[w_pos[1],w_pos[1]+w_width],[-w_width,None]]:
splits.append(data[:, :, zz[0]:zz[1], hh[0]:hh[1], ww[0]:ww[1]])
return torch.cat(splits, 0)
def combine64(output, z, h, w):
splits = []
for i in range(len(output)):
splits.append(output[i])
output = np.zeros((
z,
h,
w,
splits[0].shape[3],
splits[0].shape[4]), np.float32)
z_width = int(np.ceil(float(z) / 4))
h_width = int(np.ceil(float(h) / 4))
w_width = int(np.ceil(float(w) / 4))
splitzstart = splits[0].shape[0]/2-z_width/2
splithstart = splits[0].shape[1]/2-h_width/2
splitwstart = splits[0].shape[2]/2-w_width/2
i = 0
for zz,zz2 in zip([[0,z_width],[z_width,z_width*2],[z_width*2,z_width*3],[z_width*3-z,None]],
[[0,z_width],[splitzstart,z_width+splitzstart],[splitzstart,z_width+splitzstart],[z_width*3-z,None]]):
for hh,hh2 in zip([[0,h_width],[h_width,h_width*2],[h_width*2,h_width*3],[h_width*3-h,None]],
[[0,h_width],[splithstart,h_width+splithstart],[splithstart,h_width+splithstart],[h_width*3-h,None]]):
for ww,ww2 in zip([[0,w_width],[w_width,w_width*2],[w_width*2,w_width*3],[w_width*3-w,None]],
[[0,w_width],[splitwstart,w_width+splitwstart],[splitwstart,w_width+splitwstart],[w_width*3-w,None]]):
output[zz[0]:zz[1], hh[0]:hh[1], ww[0]:ww[1], :, :] = splits[i][zz2[0]:zz2[1], hh2[0]:hh2[1], ww2[0]:ww2[1], :, :]
i = i+1
return output
def plotlog(logfile, savepath):
traintpr = []
traintnr = []
trainloss = []
trainclassifyloss = []
trainregresslossx = []
trainregresslossy = []
trainregresslossz = []
trainregresslossd = []
valtpr = []
valtnr = []
valloss = []
valclassifyloss = []
valregresslossx = []
valregresslossy = []
valregresslossz = []
valregresslossd = []
eps = 1
f = open(logfile, 'r')
line = f.readline()
while line:
if line.startswith('Epoch '+'{:03d}'.format(eps)+' (lr '):
trainline1 = f.readline()
strlist = trainline1.split('Train: tpr ')
# print strlist
strlist1 = strlist[1].split(',')
# print strlist1, float(strlist1[0])
traintpr.append(float(strlist1[0]))
strlist2 = strlist1[1].split('tnr ')
# print strlist2, float(strlist2[1])
traintnr.append(float(strlist2[1]))
trainline2 = f.readline()[5:]
strlist = trainline2.split(', classify loss ')
# print strlist, float(strlist[0])
trainloss.append(float(strlist[0]))
strlist1 = strlist[1].split(', regress loss ')
# print strlist1, float(strlist1[0])
trainclassifyloss.append(float(strlist1[0]))
strlist2 = strlist1[1].split(', ')
# print strlist2, float(strlist2[0]), float(strlist2[1]), float(strlist2[2]), float(strlist2[3])
trainregresslossx.append(float(strlist2[0]))
trainregresslossy.append(float(strlist2[1]))
trainregresslossz.append(float(strlist2[2]))
trainregresslossd.append(float(strlist2[3]))
f.readline()
valline1 = f.readline()
strlist = valline1.split('Validation: tpr ')
# print strlist
strlist1 = strlist[1].split(',')
# print strlist1, float(strlist1[0])
valtpr.append(float(strlist1[0]))
strlist2 = strlist1[1].split('tnr ')
# print strlist2, float(strlist2[1])
valtnr.append(float(strlist2[1]))
valline2 = f.readline()[5:]
strlist = valline2.split(', classify loss ')
# print strlist, float(strlist[0])
valloss.append(float(strlist[0]))
strlist1 = strlist[1].split(', regress loss ')
# print strlist1, float(strlist1[0])
valclassifyloss.append(float(strlist1[0]))
strlist2 = strlist1[1].split(', ')
# print strlist2, float(strlist2[0]), float(strlist2[1]), float(strlist2[2]), float(strlist2[3])
valregresslossx.append(float(strlist2[0]))
valregresslossy.append(float(strlist2[1]))
valregresslossz.append(float(strlist2[2]))
valregresslossd.append(float(strlist2[3]))
eps += 1
line = f.readline()
f.close()
fig = plt.figure()
plt.plot(range(1, len(traintpr)+1, 1), traintpr, label='train')
plt.plot(range(1, len(traintpr)+1, 1), valtpr, label='val')
plt.legend()
plt.title('True Positive Rate')
plt.savefig(savepath+'tpr.png')
fig = plt.figure()
plt.plot(range(1, len(traintnr)+1, 1), traintnr, label='train')
plt.plot(range(1, len(traintpr)+1, 1), valtnr, label='val')
plt.legend()
plt.title('True Negative Rate')
plt.savefig(savepath+'tnr.png')
fig = plt.figure()
plt.plot(range(1, len(traintpr)+1, 1), trainloss, label='train')
plt.plot(range(1, len(traintpr)+1, 1), valloss, label='val')
plt.legend()
plt.title('Loss')
plt.savefig(savepath+'loss.png')
fig = plt.figure()
plt.plot(range(1, len(traintpr)+1, 1), trainclassifyloss, label='train')
plt.plot(range(1, len(traintpr)+1, 1), valclassifyloss, label='val')
plt.legend()
plt.title('Classification Loss')
plt.savefig(savepath+'classificationloss.png')
fig = plt.figure()
plt.plot(range(1, len(traintpr)+1, 1), trainregresslossx, label='train')
plt.plot(range(1, len(traintpr)+1, 1), valregresslossx, label='val')
plt.legend()
plt.title('Regresion X Loss')
plt.savefig(savepath+'regressionxloss.png')
fig = plt.figure()
plt.plot(range(1, len(traintpr)+1, 1), trainregresslossy, label='train')
plt.plot(range(1, len(traintpr)+1, 1), valregresslossy, label='val')
plt.legend()
plt.title('Regresion Y Loss')
plt.savefig(savepath+'regressionyloss.png')
fig = plt.figure()
plt.plot(range(1, len(traintpr)+1, 1), trainregresslossz, label='train')
plt.plot(range(1, len(traintpr)+1, 1), valregresslossz, label='val')
plt.legend()
plt.title('Regresion Z Loss')
plt.savefig(savepath+'regressionzloss.png')
fig = plt.figure()
plt.plot(range(1, len(traintpr)+1, 1), trainregresslossd, label='train')
plt.plot(range(1, len(traintpr)+1, 1), valregresslossd, label='val')
plt.legend()
plt.title('Regresion D Loss')
plt.savefig(savepath+'regressiondloss.png')
def plotnoduledist(annopath):
import pandas as pd
df = pd.read_csv(annopath+'train/annotations.csv')
diameter = df['diameter_mm'].reshape((-1,1))
df = pd.read_csv(annopath+'val/annotations.csv')
diameter = np.vstack([df['diameter_mm'].reshape((-1,1)), diameter])
df = pd.read_csv(annopath+'test/annotations.csv')
diameter = np.vstack([df['diameter_mm'].reshape((-1,1)), diameter])
fig = plt.figure()
plt.hist(diameter, normed=True, bins=50)
plt.ylabel('probability')
plt.xlabel('Diameters')
plt.title('Nodule Diameters Histogram')
plt.savefig('nodulediamhist.png')
def plotnoduledistkaggle(annopath):
import pandas as pd
df = pd.read_csv(annopath)
diameter = df['diameter']
dlist = []
for i in xrange(diameter.shape[0]):
if diameter[i] > 0:
dlist.append(diameter[i])
plt.hist(dlist, normed=True, bins=50)
plt.ylabel('probability')
plt.xlabel('Diameters')
plt.title('Nodule Diameters Histogram')
plt.savefig('nodulediamhistkaggle.png')
def calrecall(csvpath):
import pandas as pd
df = pd.read_csv(csvpath)
seriesid = df['seriesuid']
ttol = seriesid.shape[0]
pos = len(set(seriesid))
print(pos/float(ttol))
def plothistdiameter(trainpath='/media/data1/wentao/tianchi/preprocessing/newtrain/',
testpath='/media/data1/wentao/tianchi/preprocessing/newtest/'):
diameterlist = []
for fname in os.listdir(trainpath):
if fname.endswith('_label.npy'):
label = np.load(trainpath+fname)
for lidx in xrange(label.shape[0]):
diameterlist.append(label[lidx, -1])
for fname in os.listdir(testpath):
if fname.endswith('_label.npy'):
label = np.load(testpath+fname)
for lidx in xrange(label.shape[0]):
diameterlist.append(label[lidx, -1])
fig = plt.figure()
plt.hist(diameterlist, 50)
plt.xlabel('Nodule Diameter')
plt.ylabel('# Nodules')
plt.title('Nodule Size Histogram')
plt.savefig('processnodulesizehist.png')
def getrecall(path='/media/data1/wentao/tianchi/CTnoddetector/training/detector/results/res18/baselinebboxlranchorftbig/test55test/'):
thresh = np.linspace(-15, 0, 31)
p = 0
print thresh.shape, thresh
# print recall.size
tot = 0
for fname in os.listdir(path):
if fname.endswith('_lbb.npy'):
tot += 1
recall = np.zeros((thresh.shape[0], tot))
tot = 0
for fname in os.listdir(path):
if fname.endswith('_lbb.npy'):
print fname
label = np.load(path+fname, 'r')
p += label.shape[0]
pbb = np.load(path+fname[:-8]+'_pbb.npy', 'r')
if label.shape[0] == 0: continue
for thi in xrange(thresh.shape[0]-1, -1, -1):
if thi != thresh.shape[0]-1 and recall[thi+1, tot] == label.shape[0]:
recall[thi, tot] = label.shape[0]
continue
th = thresh[thi]
tp = 0
pbbremain = np.array(pbb[pbb[:,0]>th])
lidxbegin = 0
for pidx in xrange(pbbremain.shape[0]):
for lidx in xrange(lidxbegin, label.shape[0], 1):
if np.sum(np.square(pbbremain[pidx, 1:4] - label[lidx, :3])) <= (label[lidx,-1]/2)**2:
tp += 1
lidxbegin += 1
if lidxbegin == label.shape[0]:
break
continue
if lidxbegin == label.shape[0]:
break
recall[thi, tot] += tp
# print th, tp
tot += 1
recall /= p
print 'p', p
print recall.sum(axis=1)
fig = plt.figure()
plt.plot(thresh, recall.sum(axis=1))
plt.xlabel('Threshold (before sigmoid)')
plt.ylabel('Recall')
plt.title('Recall w.r.t. Threshold')
plt.savefig('recallthresh.png')
def getrecallwrtdetp(path='/media/data1/wentao/tianchi/CTnoddetector/training/detector/results/res18/baselinebboxlranchorftbig96/test64/'):
thresh = np.linspace(-7, 0, 8)
import pandas as pd
import math
gtcolname = ['seriesuid', 'coordX', 'coordY', 'coordZ', 'diameter_mm']
annogt = pd.read_csv('/media/data1/wentao/tianchi/csv/newtest/annotations.csv', names=gtcolname)
seriesuidgt = annogt.seriesuid.tolist()
coordXgt = annogt.coordX.tolist()
coordYgt = annogt.coordY.tolist()
coordZgt = annogt.coordZ.tolist()
diameter_mmgt = annogt.diameter_mm.tolist()
p = 0
# print recall.size
tot = 0
for fname in os.listdir(path):
if fname.endswith('_lbb.npy'):
tot += 1
label = np.load(path+fname, 'r')
p += label.shape[0]
print thresh.shape, thresh, tot, p, seriesuidgt[0]
recall = np.zeros((thresh.shape[0],))
colnames = ['seriesuid', 'coordX', 'coordY', 'coordZ', 'probability']
for thi in xrange(thresh.shape[0]):
df = pd.read_csv(path+'predanno'+'{:02d}'.format(int(thresh[thi]))+'.csv', names=colnames)
seriesuid = df.seriesuid.tolist()
coordX = df.coordX.tolist()
# print type(coordX)
coordY = df.coordY.tolist()
coordZ = df.coordZ.tolist()
# print coordX, seriesuid, coordY, coordZ
for fname in set(seriesuidgt[1:]):
labelx, labely, labelz, labelr = [], [], [], []
for seriesidx in xrange(len(seriesuidgt)):
if seriesuidgt[seriesidx] == fname:
labelx.append(float(coordXgt[seriesidx]))
labely.append(float(coordYgt[seriesidx]))
labelz.append(float(coordZgt[seriesidx]))
labelr.append(float(diameter_mmgt[seriesidx]))
labelsignlist = [0]*len(labelx)
coordXremain, coordYremain, coordZremain = [], [], []
for csvidx in xrange(len(seriesuid)):
# print seriesuid[csvidx], fname[:-8]
if seriesuid[csvidx] == fname:
coordXremain.append(coordX[csvidx])
coordYremain.append(coordY[csvidx])
coordZremain.append(coordZ[csvidx])
# print len(coordX), len(coordXremain)#, labelx#, coordXremain
for coordi in xrange(len(coordXremain)):
for labeli in xrange(len(label)):
if labelsignlist[labeli] == 1:
continue
try:
float(coordXremain[coordi])
float(coordYremain[coordi])
float(coordZremain[coordi])
except:
print coordXremain[coordi], coordYremain[coordi], coordZremain[coordi]
dist = math.pow(float(coordXremain[coordi]) - labelx[labeli], 2.0)
dist += math.pow(float(coordYremain[coordi]) - labely[labeli], 2.0)
dist += math.pow(float(coordZremain[coordi]) - labelz[labeli], 2.0)
if dist < math.pow(labelr[labeli]/2.0, 2.0):
recall[thi] += 1
labelsignlist[labeli] = 1
# break
if sum(labelsignlist) == label.shape[0]:
break
recall /= float(p)
print recall
fig = plt.figure()
plt.plot(thresh, recall)
plt.xlabel('Threshold (before sigmoid)')
plt.ylabel('Recall')
plt.title('Recall w.r.t. Threshold (after nms)')
plt.savefig('recallthreshafternms.png')
if __name__ == '__main__':
# path = '/media/data1/wentao/CTnoddetector/training/detector/results/res18/ft96/'
# plotlog(path+'log', savepath=path)
plotlog('log', savepath='./')
# plotnoduledist('/mnt/media/wentao/tianchi/csv/')
# plotnoduledistkaggle('/mnt/media/wentao/tianchi/DSB2017/training/detector/labels/label_job0_full.csv')
# calrecall('/mnt/media/wentao/tianchi/csv/val/annotations.csv')
# calrecall('/mnt/media/wentao/tianchi/csv/test/annotations.csv')
# plothistdiameter()
# getrecall()
# getrecallwrtdetp()
