我似乎总是陷入困境,迷失在如何重塑数据以适应模型。我认为输入和输出数据的形状必须匹配,但我一直迷失在如何解决这个问题上。
我认为我的主要问题是灰度图像和RGB图像的存储方式不同。 [1] vs [255,255,255]
所以如果:
screen = cv2.cvtColor(screen,cv2.COLOR_BGR2RGB)
更改为:
screen = cv2.cvtColor(screen,cv2.COLOR_BGR2GRAY)
一切正常。
有问题的代码:
# Capture Data (CUT SHORT)
WIDTH = 160
HEIGHT = 120
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
screen = cv2.resize(screen, (WIDTH, HEIGHT))
dataset = []
output = [0, 0, 0, 0]
dataset.append([screen, output])
np.save("training.npy", dataset)
# Build Model
https://github.com/tflearn/tflearn/blob/master/examples/images/alexnet.py
# Changed to match output.
network = fully_connected(network, 4, activation='softmax')
# Train Data
WIDTH = 160
HEIGHT = 120
LR = 1e-3
EPOCHS = 5
MODEL_NAME = "HELP"
model = alexnet(WIDTH, HEIGHT, LR)
for i in range(EPOCHS):
train_data = np.load("training.npy".format(i))
train = train_data[:-100]
test = train_data[-100:]
X = np.array([i[0] for i in train]).reshape(-1,WIDTH,HEIGHT,1)
Y = [i[1] for i in train]
test_x = np.array([i[0] for i in test]).reshape(-1,WIDTH,HEIGHT,1)
test_y = [i[1] for i in test]
model.fit({'input': X}, {'targets': Y}, n_epoch=1, validation_set=({'input': test_x}, {'targets': test_y}),
snapshot_step=500, show_metric=True, run_id=MODEL_NAME)
model.save(MODEL_NAME)
错误: 线程Thread-3中的异常: Traceback(最近一次调用最后一次): 文件" C:\ Users \ TF \ AppData \ Local \ Programs \ Python \ Python35 \ lib \ threading.py",第914行,在_bootstrap_inner中 self.run() 文件" C:\ Users \ TF \ AppData \ Local \ Programs \ Python \ Python35 \ lib \ threading.py",第862行,运行中 self._target(* self._args,** self._kwargs) 文件" C:\ Users \ TF \ AppData \ Local \ Programs \ Python \ Python35 \ lib \ site-packages \ tflearn \ data_flow.py",第187行,在fill_feed_dict_queue中 data = self.retrieve_data(batch_ids) 在retrieve_data中的文件" C:\ Users \ TF \ AppData \ Local \ Programs \ Python \ Python35 \ lib \ site-packages \ tflearn \ data_flow.py",第222行 utils.slice_array(self.feed_dict [key],batch_ids) 文件" C:\ Users \ TF \ AppData \ Local \ Programs \ Python \ Python35 \ lib \ site-packages \ tflearn \ utils.py",第187行,在slice_array中 返回X [开始]
IndexError:索引2936超出了轴0的大小为1900
的范围答案 0 :(得分:0)
博士。 Robert Kirchgessner: 输入数据集中有三个通道。
np.array([i[0] for i in test]).reshape(-1,WIDTH,HEIGHT,3)
在alexnet:
network = input_data(shape=[None, width, height, 3], name='input')