def try_different_method(clf, x_train, y_train, x_test, y_test):
clf.fit(x_train, y_train)
score = clf.score(x_test, y_test)
print('score: ', score)
result = clf.predict(x_test)
plt.figure()
plt.plot(np.arange(len(result)), y_test, 'go-', label='true value')
plt.plot(np.arange(len(result)), result, 'ro-', label='predict value')
plt.title('score: %f' % score)
plt.legend()
plt.show()
# 保存Model
with open('./clf.pickle', 'wb') as f:
pickle.dump(clf, f)
# 读取Model
with open('./clf.pickle', 'rb') as f:
clf2 = pickle.load(f)
# 测试读取后的Model
print(clf2.predict(x_test[0:1]), y_test[0:1])
print('relative x error: ', x_err_esti(result, y_test))
print('mean absolute error: ', vx_err_esti(result, y_test))
result = clf2.predict([x_test[0]])
print(x_test[0])
print(result)
def train_model(clf, t_smples, t_targets):
x_train, x_test, y_train, y_test = train_test_split(t_smples,
t_targets,
test_size=0.1,
random_state=42)
clf.fit(x_train, y_train)
result = clf.predict(x_test)
print('Train relative x error: ', x_err_esti(result, y_test))
# print('Train mean absolute error: ', vx_err_esti(result, y_test))
return clf
def test(x_test, y_test):
# 读取Model
with open('./clf.pickle', 'rb') as f:
clf = pickle.load(f)
# 测试读取后的Model
print(clf.predict(x_test[0:1]), y_test[0:1])
result = clf.predict(x_test)
print('relative x error: ', x_err_esti(result, y_test))
print('mean absolute error: ', vx_err_esti(result, y_test))
def err_estimation(self, gt_file, before):
result = self.result
fids = [r['fid'] for r in result]
vx = [r['vx'] for r in result]
x = [r['x'] for r in result]
with open(gt_file) as f:
ground_truth = json.load(f)
gt_x = [gt['x'] for gt in ground_truth["frame_data"]]
gt_vx = [gt['vx'] for gt in ground_truth["frame_data"]]
# before = 10
err_x = x_err_esti(x[:before], gt_x[:before])
err_vx = vx_err_esti(vx[:before], gt_vx[:before])
high = high_esti(x[:before], gt_x[:before])
print('Error for x :', err_x)
print('Error for vx:', err_vx)
print('high :', high)
def err(prediction, gt_x):
prediction = np.asarray(prediction, np.float32)
gt_x = np.asarray(gt_x, np.float32)
return x_err_esti(prediction, gt_x)
X_test, y_test, M, ind)
print("test video :", test_flies[ind])
feed_dict = {
model.xs: X_test_batch,
# model.cell_init_state: state # use last state as the initial state for this run
}
y_test_pred, outputs = sess.run(
[model.pred, model.cell_outputs], feed_dict=feed_dict)
# print('*'*50+'outputs'+'*'*50)
# print(outputs)
xs = np.arange(500)
if x_vx_mode == 'x':
print(
'error for x:',
x_err_esti(y_test_pred[:, :, 0], y_test_batch[:, :,
0]))
draw_line(xs,
y_test_pred,
y_test_batch,
name='show/curve_train/x_%s.png' %
test_flies[ind][:-7])
elif x_vx_mode == 'vx':
print(
'error for vx:',
vx_err_esti(y_test_pred[:, :, 0],
y_test_batch[:, :, 0]))
draw_line(xs,
y_test_pred,
y_test_batch,
name='show/curve_train/vx_%s.png' %
test_flies[ind][:-7])
def test_model(clf, x_test, y_test):
result = clf.predict(x_test)
print('Test relative x error: ', x_err_esti(result, y_test))