python - (0)未知: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize

Question

在这里，我使用mask_rcnn进行对象检测。
代码运行正常，没有错误，但是现在出现以下错误。

代码:

Testing - 1
# 1 - 200 images testing
# Load a random image from the images folder
file_names = next(os.walk(IMAGE_DIR))[2]
i=0


for fn in file_names:

    #image = skimage.io.imread(os.path.join(IMAGE_DIR, random.choice(file_names)))
    i = i+1
    print(i)

    image = skimage.io.imread(os.path.join(IMAGE_DIR, fn))

    # Run detection
    results = model.detect([image], verbose=1)

    # Visualize results
    r = results[0]

    visualize.display_instances(image, r['rois'], r['masks'], r['class_ids'], 
                            class_names, r['scores'])

错误信息 :
1
处理1张图像
图像形状:(1123，1588，3)最小值:13.00000最大值:255.00000 uint8
moulded_images形状:(1、1024、1024、3)最小值:-123.70000最大值:150.10000 float64
image_metas形状:(1，15)最小值:0.00000最大值:1588.00000 float64
anchor 定形状:(1，261888，4)最小值:-0.35390最大值:1.29134 float32
-------------------------------------------------- -------------------------
UnknownError跟踪(最近一次通话)
在
15
16＃运行检测
---> 17个结果= model.detect([image]，verbose = 1)
18岁
19＃可视化结果

D:\gpu\projects1\Mask_RCNN-Multi-Class-Detection\model.py in detect(self, images, verbose)
   2529         # Run object detection
   2530         detections, _, _, mrcnn_mask, _, _, _ =\
-> 2531             self.keras_model.predict([molded_images, image_metas, anchors], verbose=0)
   2532         # Process detections
   2533         results = []

~\Anaconda3\envs\tf-gpu\lib\site-packages\keras\engine\training.py in predict(self, x, batch_size, verbose, steps, callbacks, max_queue_size, workers, use_multiprocessing)
   1399                                             verbose=verbose,
   1400                                             steps=steps,
-> 1401                                             callbacks=callbacks)
   1402 
   1403     def train_on_batch(self, x, y,

~\Anaconda3\envs\tf-gpu\lib\site-packages\keras\engine\training_arrays.py in predict_loop(model, f, ins, batch_size, verbose, steps, callbacks)
    330             batch_logs = {'batch': batch_index, 'size': len(batch_ids)}
    331             callbacks._call_batch_hook('predict', 'begin', batch_index, batch_logs)
--> 332             batch_outs = f(ins_batch)
    333             batch_outs = to_list(batch_outs)
    334             if batch_index == 0:

~\Anaconda3\envs\tf-gpu\lib\site-packages\keras\backend\tensorflow_backend.py in __call__(self, inputs)
   2977                     return self._legacy_call(inputs)
   2978 
-> 2979             return self._call(inputs)
   2980         else:
   2981             if py_any(is_tensor(x) for x in inputs):

~\Anaconda3\envs\tf-gpu\lib\site-packages\keras\backend\tensorflow_backend.py in _call(self, inputs)
   2935             fetched = self._callable_fn(*array_vals, run_metadata=self.run_metadata)
   2936         else:
-> 2937             fetched = self._callable_fn(*array_vals)
   2938         return fetched[:len(self.outputs)]
   2939 

~\Anaconda3\envs\tf-gpu\lib\site-packages\tensorflow\python\client\session.py in __call__(self, *args, **kwargs)
   1456         ret = tf_session.TF_SessionRunCallable(self._session._session,
   1457                                                self._handle, args,
-> 1458                                                run_metadata_ptr)
   1459         if run_metadata:
   1460           proto_data = tf_session.TF_GetBuffer(run_metadata_ptr)

UnknownError: 2 root error(s) found.
  (0) Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
     [[{{node conv1/convolution}}]]
     [[mrcnn_detection/map/TensorArrayUnstack/range/_4207]]
  (1) Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
     [[{{node conv1/convolution}}]]
0 successful operations.
0 derived errors ignored.

Answer 1

您必须检查CUDA + CUDNN + TensorFlow的版本是否正确(还要确保已安装所有版本。

下面提供了两个运行配置示例

例如。 Cuda 10 + CuDNN 7.6.3 + TensorFlow 1.13/1.14运行。

Eg2 Cuda 9 + CuDNN 7.0.5 + TensorFlow 1.10工作

Eg3 Cuda 10.1 + CuDNN 7.6.5 + TensorFlow 2.1作品

通常，当您安装了不兼容的TensorFlow/CuDNN版本时，会出现此错误。就我而言，这是当我尝试使用较旧的TensorFlow和较新版本的CuDNN时出现的。