faster_rcnn_r50_1x 训练完成，模型已导出，但在预测时报错。

[wyc880622]

模型导出应该没有问题。
报错内容如下所示：请问大师该如何解决？

(paddle) G:\Halcon\Paddle\PaddleDetection>python -u tools/infer.py -c configs/faster_rcnn_r50_1x.yml -o weights=inference_model\faster_rcnn_r50_1x --infer_img=demo/1.jpg --output_dir=infer_output
W0423 19:38:42.788797 21428 device_context.cc:237] Please NOTE: device: 0, CUDA Capability: 61, Driver API Version: 10.1, Runtime API Version: 10.0
W0423 19:38:42.800806 21428 device_context.cc:245] device: 0, cuDNN Version: 7.6.
2020-04-23 19:38:44,095-INFO: Loading parameters from inference_model\faster_rcnn_r50_1x...
2020-04-23 19:38:44,095-WARNING: inference_model\faster_rcnn_r50_1x.pdparams not found, try to load model file saved with [ save_params, save_persistables, save_vars ]
2020-04-23 19:38:44,095-WARNING: inference_model\faster_rcnn_r50_1x.pdparams not found, try to load model file saved with [ save_params, save_persistables, save_vars ]
2020-04-23 19:38:44,098-WARNING: variable file [ inference_model/faster_rcnn_r50_1x/params inference_model/faster_rcnn_r50_1x/model ] not used
2020-04-23 19:38:44,098-WARNING: variable file [ inference_model/faster_rcnn_r50_1x/params inference_model/faster_rcnn_r50_1x/model ] not used
C:\Users\Administrator\Anaconda3\envs\paddle\lib\site-packages\paddle\fluid\executor.py:804: UserWarning: There are no operators in the program to be executed. If you pass Program manually, please use fluid.program_guard to ensure the current Program is being used.
warnings.warn(error_info)
2020-04-23 19:38:44,100-INFO: Load categories from G:/Halcon/Paddle/PaddleDetection/dataset/coco/annotations/instance_test.json
loading annotations into memory...
Done (t=0.00s)
creating index...
index created!
C:\Users\Administrator\Anaconda3\envs\paddle\lib\site-packages\paddle\fluid\executor.py:782: UserWarning: The following exception is not an EOF exception.
"The following exception is not an EOF exception.")
Traceback (most recent call last):
File "tools/infer.py", line 271, in
main()
File "tools/infer.py", line 185, in main
return_numpy=False)
File "C:\Users\Administrator\Anaconda3\envs\paddle\lib\site-packages\paddle\fluid\executor.py", line 783, in run
six.reraise(*sys.exc_info())
File "C:\Users\Administrator\Anaconda3\envs\paddle\lib\site-packages\six.py", line 703, in reraise
raise value
File "C:\Users\Administrator\Anaconda3\envs\paddle\lib\site-packages\paddle\fluid\executor.py", line 778, in run
use_program_cache=use_program_cache)
File "C:\Users\Administrator\Anaconda3\envs\paddle\lib\site-packages\paddle\fluid\executor.py", line 831, in _run_impl
use_program_cache=use_program_cache)
File "C:\Users\Administrator\Anaconda3\envs\paddle\lib\site-packages\paddle\fluid\executor.py", line 902, in _run_program
self._feed_data(program, feed, feed_var_name, scope)
File "C:\Users\Administrator\Anaconda3\envs\paddle\lib\site-packages\paddle\fluid\executor.py", line 580, in _feed_data
check_feed_shape_type(var, cur_feed)
File "C:\Users\Administrator\Anaconda3\envs\paddle\lib\site-packages\paddle\fluid\executor.py", line 230, in check_feed_shape_type
(var.name, len(var.shape), var.shape, feed_shape))
ValueError: The fed Variable 'image' should have dimensions = 4, shape = (-1, 3, 800, 1333), but received fed shape [1, 3, 800, 1067] on each device

[jerrywgz]

这个是因为paddle比较新的版本会对网络定义shape和实际输入shape做检查，可以更新下检测库，release/0.2分支就已经适配了

[yghstill]

看您使用，还有问题就是导出的inference_model模型请使用INFERENCE方法进行预测。不然加载不了模型，检测不到结果。
如果想要使用tools/infer.py进行预测，不用导出模型，请直接使用训练完成的模型进行预测即可。

[wyc880622]

release/0.2分支就已经适配了

谢谢你的回答，我在用的就是0.2版本paddleDetection。是和你说的是一个意思吗。

[jerrywgz]

可以看下你的配置文件吗，另外确实需要再确认下你infer时加载的模型是否是训练时保存的模型

[wyc880622]

配置文件如下：
configs.zip
infer加载的模型是导出的模型。但我 用直接导出的模型预测，也同样是报一样的错误。现在不知道应该如何处理了。谢谢

[wyc880622]

这个文件是 导出模型和没有导出模型预测，结果都是eof错误。我赢怎么更新下检测库

[jerrywgz]

对于infer.py的问题，看了下你的配置文件，你在faster_reader.yml中的配置和https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.2/configs/faster_reader.yml 这里的配置还有些出入，这个也可能是release/0.2期间有过更新，你本地可以再更新下这个分支，
主要是要将配置文件中红框的部分去掉才能正常预测

关于cpp_infer的问题，是由于cpp_demo这个配置文件是使用rcnn系列模型的，你这边用的是yolo的模型，这两个模型在输入部分会有些出入，这个我推荐使用master分支，export_model.py会自动根据你测试时的配置生成cpp_infer所需要的配置文件

[wyc880622]

我已将yml进行了修改，进行了推理程序没有错误。但是预测输出的图片并没有框上，我不知道为什么，难道深度学习训练文件出现了问题，我感觉不太可能，怎么也可以识别出1个来。

请问我还应该怎么解决？

[yghstill]

@wyc880622 请使用--draw_threshold参数调小可视化得分阈值，如果阈值很小也检测不到，那就是训练模型出了问题。另外，是什么模型，loss收敛到什么程度，用训练集图像测试也检测不到吗？

[wyc880622]

谢谢你回复，我map=0 应该是没有训练好，后来我又从新训练，在保存节点模型时报错，报错内容如下，我训练了2次都出现了这种情况。
请问是哪里的错误？

2020-04-24 18:24:58,539-INFO: iter: 9980, lr: 0.010000, 'loss_cls': '0.003621',
'loss_bbox': '0.000801', 'loss_rpn_cls': '0.000050', 'loss_rpn_bbox': '0.000137'
, 'loss': '0.004480', time: 1.147, eta: 1 day, 4:40:48
2020-04-24 18:25:21,467-INFO: iter: 10000, lr: 0.010000, 'loss_cls': '0.002752',
'loss_bbox': '0.000784', 'loss_rpn_cls': '0.000061', 'loss_rpn_bbox': '0.000090
', 'loss': '0.004001', time: 1.147, eta: 1 day, 4:40:07
2020-04-24 18:25:21,468-INFO: Save model to output\faster_rcnn_r50_1x\10000.
I0424 18:25:24.260897 5308 parallel_executor.cc:440] The Program will be execut
ed on CUDA using ParallelExecutor, 1 cards are used, so 1 programs are executed
in parallel.
I0424 18:25:24.267897 5308 build_strategy.cc:365] SeqOnlyAllReduceOps:0, num_tr
ainers:1
2020-04-24 18:25:24,945-INFO: Test iter 0
2020-04-24 18:25:50,377-INFO: Test finish iter 42
2020-04-24 18:25:50,378-INFO: Total number of images: 42, inference time: 1.6077
78591672592 fps.
loading annotations into memory...
Done (t=0.00s)
creating index...
index created!
2020-04-24 18:25:50,383-INFO: Start evaluate...
Loading and preparing results...
DONE (t=0.00s)
creating index...
index created!
Traceback (most recent call last):
File "C:\ProgramData\Anaconda3\envs\paddle\lib\site-packages\numpy\core\functi
on_base.py", line 117, in linspace
num = operator.index(num)
TypeError: 'numpy.float64' object cannot be interpreted as an integer

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "tools/train.py", line 331, in
main()
File "tools/train.py", line 274, in main
cfg['EvalReader']['dataset'])
File "E:\paddle\PaddleDetection-release-0.2\ppdet\utils\eval_utils.py", line 2
09, in eval_results
is_bbox_normalized=is_bbox_normalized)
File "E:\paddle\PaddleDetection-release-0.2\ppdet\utils\coco_eval.py", line 94
, in bbox_eval

[yghstill]

@wyc880622 由于numpy的最新1.18版本改了一个接口的参数类型，pycocotools里评估时用到这个接口，就报错了，请回退numpy版本或者升级pycocotools，具体参考 #245 #445 的issue解决方案。

[yghstill]

@wyc880622 如果没有问题先关闭了，还有问题的话可以reopen这个issue。