def test_cnn_val():
dataset_params = {
'batch_size': -1,
'path': 'resources/train_data/chars',
'labels_path': 'resources/train_data/chars_list_val.pickle',
'thread_num': 3,
'gray': True
}
val_dataset_reader = DataSet(dataset_params)
model = Lenet()
model.compile()
image, label = val_dataset_reader.batch()
print("Total dataset number: ", val_dataset_reader.record_number)
pred, acc = eval_model([model.pred_labels, model.accuracy], {
model.x: image,
model.y: label,
model.keep_prob: 1
},
model_dir="train/model/chars/models/")
print("Label: {}({}), Pred: {}({})".format(label[0], index2str[label[0]],
pred[0], index2str[pred[0]]))
# imshow("tmp", image[0])
print("Accuary: {:.2f}%".format(acc * 100))
def test_judge_val():
dataset_params = {
'batch_size': -1,
'path': 'resources/train_data/whether_car',
'labels_path': 'resources/train_data/whether_list_val.pickle',
'thread_num': 3,
'gray': False
}
val_dataset_reader = DataSet(dataset_params)
model = Judgenet()
model.compile()
image, label = val_dataset_reader.batch()
print("Total dataset number: ", val_dataset_reader.record_number)
pred, acc = eval_model([model.pred_labels, model.accuracy], {
model.x: image,
model.y: label,
model.keep_prob: 1
},
model_dir="train/model/whether_car/models/")
print("Label: {}, Pred: {}".format(label[0], pred[0]))
print("Accuary: {:.2f}%".format(acc * 100))
def identify(self, images):
tmp = images / 255 * 2 - 1
pred = eval_model(self.model.pred_labels, {
self.model.x: tmp,
self.model.keep_prob: 1
},
model_dir="train/model/chars/models/",
eval_sess=self.eval_sess)
return pred
def judge(self, images):
judgeRes = []
tmp = images / 255 * 2 - 1
pred = eval_model(self.model.pred_labels, {
self.model.x: tmp,
self.model.keep_prob: 1
},
model_dir="train/model/whether_car/models/",
eval_sess=self.eval_sess)
for i, tmp in enumerate(pred):
if tmp == 1:
judgeRes.append(images[i])
return judgeRes