Keras-yolo3: NameError: name 'yolo_head' is not defined

[JudyMRSD]

When I save weights during training using your original code, I got:

InvalidArgumentError (see above for traceback): Input to reshape is a tensor with 12675 values, but the requested shape requires a multiple of 3042
     [[Node: Reshape_3 = Reshape[T=DT_FLOAT, Tshape=DT_INT32, _device="/job:localhost/replica:0/task:0/device:CPU:0"](conv2d_59/BiasAdd, Reshape_3/shape)]]

Does this mean I should save model instead of weights? May I ask why in your code "model.save_weights" is used instead of "model.save_path"?

When I saved checkpoints in train.py as keras model, I wasn't able to use it in yolo.py using load_model.

checkpoint = ModelCheckpoint(log_dir + "checkpoint.h5", monitor='val_loss', save_best_only=True)

history = model.fit([image_data, *y_true],
          np.zeros(len(image_data)),
          validation_split=.1,
          batch_size=Batch_Size,
          epochs=10000,
          callbacks=[checkpoint])

When I use yolo.py to test the trained model on images, model is loaded using

self.yolo_model = load_model(model_path, compile=False)

I got this error:

2018-05-10 05:47:25.060203: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 FMA
Traceback (most recent call last):
  File "/home/jin/workspace/Intersection_TrafficFlow/detection/keras-yolo3-noaug/keras-yolo3-1/yolo.py", line 229, in <module>
    detect_img(YOLO())
  File "/home/jin/workspace/Intersection_TrafficFlow/detection/keras-yolo3-noaug/keras-yolo3-1/yolo.py", line 58, in __init__
    self.boxes, self.scores, self.classes = self.generate()
  File "/home/jin/workspace/Intersection_TrafficFlow/detection/keras-yolo3-noaug/keras-yolo3-1/yolo.py", line 80, in generate
    self.yolo_model = load_model(model_path, compile=False)
  File "/home/jin/anaconda3/envs/keras-yolo3-noaug/lib/python3.6/site-packages/keras/models.py", line 243, in load_model
    model = model_from_config(model_config, custom_objects=custom_objects)
  File "/home/jin/anaconda3/envs/keras-yolo3-noaug/lib/python3.6/site-packages/keras/models.py", line 317, in model_from_config
    return layer_module.deserialize(config, custom_objects=custom_objects)
  File "/home/jin/anaconda3/envs/keras-yolo3-noaug/lib/python3.6/site-packages/keras/layers/__init__.py", line 55, in deserialize
    printable_module_name='layer')
  File "/home/jin/anaconda3/envs/keras-yolo3-noaug/lib/python3.6/site-packages/keras/utils/generic_utils.py", line 144, in deserialize_keras_object
    list(custom_objects.items())))
  File "/home/jin/anaconda3/envs/keras-yolo3-noaug/lib/python3.6/site-packages/keras/engine/topology.py", line 2524, in from_config
    process_node(layer, node_data)
  File "/home/jin/anaconda3/envs/keras-yolo3-noaug/lib/python3.6/site-packages/keras/engine/topology.py", line 2483, in process_node
    layer(input_tensors, **kwargs)
  File "/home/jin/anaconda3/envs/keras-yolo3-noaug/lib/python3.6/site-packages/keras/engine/topology.py", line 619, in __call__
    output = self.call(inputs, **kwargs)
  File "/home/jin/anaconda3/envs/keras-yolo3-noaug/lib/python3.6/site-packages/keras/layers/core.py", line 663, in call
    return self.function(inputs, **arguments)
  File "/home/jinz1/Jin/Intersection_TrafficFlow/detection/keras-yolo3-noaug/keras-yolo3-1/yolo3/model.py", line 347, in yolo_loss
NameError: name 'yolo_head' is not defined

Thank you very much for your help and for creating this repo!

[qqwweee]

With weight files, you can run model = yolo_body(Input(shape=(None, None, 3)), 3, num_classes) to create model structure, then model.load_weights('weights.h5') to load weights.

Your error 1, the model handles 20 classes, but you want it to handle 1. There is a conflict.
Error 2, yolo_head is in yolo3/model.py, so the error shouldn't happen.

[zhaolei-momo]

Error 2, yolo_head is in yolo3/model.py, so the error shouldn't happen.

I got this error too.

[arvindhbti]

Your error 1, the model handles 20 classes, but you want it to handle 1.
I got this error too but I am not able to see that where to watch this classes bcz I have given correct classes

[zhiyoujingtian]

you can

model = load_model('smoke_final.h5',{'yolo_head': yolo_head})

but is other error

  File "D:\yolov3\keras-yolo3_phone\yolo3\model.py", line 376, in yolo_loss
    anchors[anchor_mask[l]], num_classes, input_shape, calc_loss=True)
TypeError: list indices must be integers or slices, not list

[Michael-Jing]

Error 2, yolo_head is in yolo3/model.py, so the error shouldn't happen.

I got this error too.

Yes, me too, and I can confirm that yolo_head is there right in the file, this is tricky.

[Michael-Jing]

I managed to get rid of the error without modifying parameters to the load_model function. In my case, the reason for the issue is that the classes definition file is different during training and testing. Not sure why this would cause python to complain that yolo_head is not defined.

[vector-1127]

I got this error too. @qqwweee How can we get rid of the error? Thanks.

[cguentherTUChemnitz]

you can

model = load_model('smoke_final.h5',{'yolo_head': yolo_head})

but is other error

  File "D:\yolov3\keras-yolo3_phone\yolo3\model.py", line 376, in yolo_loss
    anchors[anchor_mask[l]], num_classes, input_shape, calc_loss=True)
TypeError: list indices must be integers or slices, not list

I stuck with exactly this error when trying to load an model which is saved with model.save instead of model.save_weights.

[vector-1127]

Excitedly, I have solved the problem.

The core problem is "the trained model that you want to save and the model defined in the ---.cfg file must be same".

Nacturally, during training, the model has only one output in this code, i.e., L2 loss. However, in the ---.cfg file, the output is a group of feature maps that is determined by the number of classes and the number of anchors.

So, the solution is that "First, defining a network that is same with the model in the ---.cfg file. Then, loading the learned parameters to this defined network and using 'model.save('model.h5')' to save this model. Final, convert it to any form".

This way has been verified on the task of converting a keras model (.h5) to a darknet model.

[qianliangchang]

Excitedly, I have solved the problem.

The core problem is "the trained model that you want to save and the model defined in the ---.cfg file must be same".

Nacturally, during training, the model has only one output in this code, i.e., L2 loss. However, in the ---.cfg file, the output is a group of feature maps that is determined by the number of classes and the number of anchors.

So, the solution is that "First, defining a network that is same with the model in the ---.cfg file. Then, loading the learned parameters to this defined network and using 'model.save('model.h5')' to save this model. Final, convert it to any form".

This way has been verified on the task of converting a keras model (.h5) to a darknet model.

hello，I also got the issue when convert "yolo.h5" --> "yolo.pb". Follow your suggestions,

1. I have change model.save_weights() to model.save(),
2. but I don't understand the ---.cfg that above-mentionded , do you means the file yolov3.cfg that in the project?

The error information as follow:
layer(input_tensors, **kwargs) File "/home/ql/anaconda3/envs/tf/lib/python3.5/site-packages/keras/engine/topology.py", line 619, in __call__ output = self.call(inputs, **kwargs) File "/home/ql/anaconda3/envs/tf/lib/python3.5/site-packages/keras/layers/core.py", line 663, in call return self.function(inputs, **arguments) File "/home/ql/work/projects/keras-yolo3/yolo3/model.py", line 375, in yolo_loss grid, raw_pred, pred_xy, pred_wh = yolo_head(yolo_outputs[l], NameError: name 'yolo_head' is not defined
Thanks for your help!

[vector-1127]

Excitedly, I have solved the problem.
The core problem is "the trained model that you want to save and the model defined in the ---.cfg file must be same".
Nacturally, during training, the model has only one output in this code, i.e., L2 loss. However, in the ---.cfg file, the output is a group of feature maps that is determined by the number of classes and the number of anchors.
So, the solution is that "First, defining a network that is same with the model in the ---.cfg file. Then, loading the learned parameters to this defined network and using 'model.save('model.h5')' to save this model. Final, convert it to any form".
This way has been verified on the task of converting a keras model (.h5) to a darknet model.

hello，I also got the issue when convert "yolo.h5" --> "yolo.pb". Follow your suggestions,

1. I have change model.save_weights() to model.save(),
2. but I don't understand the ---.cfg that above-mentionded , do you means the file yolov3.cfg that in the project?

The error information as follow:

  layer(input_tensors, **kwargs)
 File "/home/ql/anaconda3/envs/tf/lib/python3.5/site-packages/keras/engine/topology.py", line 619, in __call__
   output = self.call(inputs, **kwargs)
 File "/home/ql/anaconda3/envs/tf/lib/python3.5/site-packages/keras/layers/core.py", line 663, in call
   return self.function(inputs, **arguments)
 File "/home/ql/work/projects/keras-yolo3/yolo3/model.py", line 375, in yolo_loss
   grid, raw_pred, pred_xy, pred_wh = yolo_head(yolo_outputs[l],
NameError: name 'yolo_head' is not defined

Thanks for your help!

Yes.

[qianliangchang]

Yes.

but I never change the yolov3.cfg, so I think the model saved same with the yolov3.cfg, any suggestions would you provide?

[qianliangchang]

Yes.

and I use this script convert yolo.h5 to yolo.pb, any clue would find? Thanks very much!
https://github.com/amir-abdi/keras_to_tensorflow/issues

[msnh2012]

@vector-1127 哥们，你的意思是要用tensorflow按照---.cfg文件自己把Yolov3构建出来么？

[vector-1127]

@vector-1127 哥们，你的意思是要用tensorflow按照---.cfg文件自己把Yolov3构建出来么？

对的，因为保存的模型要和cfg中的模型一样。

[msnh2012]

@vector-1127 感谢这么快的回复，能给个DEMO么，万分感谢

[ZH-Lee]

@vector-1127 哥们，你的意思是要用tensorflow按照---.cfg文件自己把Yolov3构建出来么？

对的，因为保存的模型要和cfg中的模型一样。

请问，这个怎么构建啊？那作者训练好的是不是就是修改过cfg的？

[ZH-Lee]

@vector-1127 感谢这么快的回复，能给个DEMO么，万分感谢
好像是你必须训练结束要保存权重文件.weights结尾，然后构建cfg，然后用convert.py 来转换

[ZH-Lee]

Excitedly, I have solved the problem.

The core problem is "the trained model that you want to save and the model defined in the ---.cfg file must be same".

Nacturally, during training, the model has only one output in this code, i.e., L2 loss. However, in the ---.cfg file, the output is a group of feature maps that is determined by the number of classes and the number of anchors.

So, the solution is that "First, defining a network that is same with the model in the ---.cfg file. Then, loading the learned parameters to this defined network and using 'model.save('model.h5')' to save this model. Final, convert it to any form".

This way has been verified on the task of converting a keras model (.h5) to a darknet model.

按道理来说，我们的这个模型都没改过，为什么要自己重新构建cfg呢？应该是一样的啊

[mengpengfei]

any clue would find

你用这个脚本转换成功了吗？

https://github.com/amir-abdi/keras_to_tensorflow

do you convert the .h5 model to .pb model successfully with above script

[sirius0503]

With weight files, you can run model = yolo_body(Input(shape=(None, None, 3)), 3, num_classes) to create model structure, then model.load_weights('weights.h5') to load weights.

Your error 1, the model handles 20 classes, but you want it to handle 1. There is a conflict.
Error 2, yolo_head is in yolo3/model.py, so the error shouldn't happen.

@qqwweee : But, the error 2 is occuring, even I've changed the number of classes to 1, by having a new classes.txt file with only 1 class.

[Rainweic]

@vector-1127 哥们，你的意思是要用tensorflow按照---.cfg文件自己把Yolov3构建出来么？

对的，因为保存的模型要和cfg中的模型一样。

刚从pytorch转tf。。。没理解要怎么做，能给个教学么 谢谢！卡在这里好久了

[Rainweic]

Excitedly, I have solved the problem.
The core problem is "the trained model that you want to save and the model defined in the ---.cfg file must be same".
Nacturally, during training, the model has only one output in this code, i.e., L2 loss. However, in the ---.cfg file, the output is a group of feature maps that is determined by the number of classes and the number of anchors.
So, the solution is that "First, defining a network that is same with the model in the ---.cfg file. Then, loading the learned parameters to this defined network and using 'model.save('model.h5')' to save this model. Final, convert it to any form".
This way has been verified on the task of converting a keras model (.h5) to a darknet model.

hello，I also got the issue when convert "yolo.h5" --> "yolo.pb". Follow your suggestions,

1. I have change model.save_weights() to model.save(),
2. but I don't understand the ---.cfg that above-mentionded , do you means the file yolov3.cfg that in the project?

The error information as follow:

  layer(input_tensors, **kwargs)
 File "/home/ql/anaconda3/envs/tf/lib/python3.5/site-packages/keras/engine/topology.py", line 619, in __call__
   output = self.call(inputs, **kwargs)
 File "/home/ql/anaconda3/envs/tf/lib/python3.5/site-packages/keras/layers/core.py", line 663, in call
   return self.function(inputs, **arguments)
 File "/home/ql/work/projects/keras-yolo3/yolo3/model.py", line 375, in yolo_loss
   grid, raw_pred, pred_xy, pred_wh = yolo_head(yolo_outputs[l],
NameError: name 'yolo_head' is not defined

Thanks for your help!

哥 解决了么 能教教我么 我也是在转化过程遇到这个Error

[Rainweic]

you can

model = load_model('smoke_final.h5',{'yolo_head': yolo_head})

but is other error

  File "D:\yolov3\keras-yolo3_phone\yolo3\model.py", line 376, in yolo_loss
    anchors[anchor_mask[l]], num_classes, input_shape, calc_loss=True)
TypeError: list indices must be integers or slices, not list

哥，我用你的方法进行测试

from tensorflow import keras 
from yolo3.model import yolo_head

model = keras.models.load_model('./trained_final.h5', {'yolo_head': yolo_head})

但是 谜一样的报错出现了：

  File "/home/keras-YOLOv3-mobilenet/yolo3/model.py", line 386, in yolo_loss
NameError: name 'tf' is not defined

没有改动过model.py文件，并且也明确了有 import tensorflow as tf
能帮忙看看么 谢谢

[Terrence1]

Input

With weight files, you can run model = yolo_body(Input(shape=(None, None, 3)), 3, num_classes) to create model structure, then model.load_weights('weights.h5') to load weights.

Your error 1, the model handles 20 classes, but you want it to handle 1. There is a conflict.
Error 2, yolo_head is in yolo3/model.py, so the error shouldn't happen.

Hi, So what is the meaning of Input and num_classes, how to define it ? thanks.

[pacelu]

issues

兄弟解决了没

[Gannyn]

you can

model = load_model('smoke_final.h5',{'yolo_head': yolo_head})

but is other error

  File "D:\yolov3\keras-yolo3_phone\yolo3\model.py", line 376, in yolo_loss
    anchors[anchor_mask[l]], num_classes, input_shape, calc_loss=True)
TypeError: list indices must be integers or slices, not list

哥，我用你的方法进行测试

from tensorflow import keras 
from yolo3.model import yolo_head

model = keras.models.load_model('./trained_final.h5', {'yolo_head': yolo_head})

但是 谜一样的报错出现了：

  File "/home/keras-YOLOv3-mobilenet/yolo3/model.py", line 386, in yolo_loss
NameError: name 'tf' is not defined

没有改动过model.py文件，并且也明确了有 import tensorflow as tf
能帮忙看看么 谢谢

请问你解决这个问题了吗？我也遇到一样的问题

[krishnak]

Excitedly, I have solved the problem.

The core problem is "the trained model that you want to save and the model defined in the ---.cfg file must be same".

Nacturally, during training, the model has only one output in this code, i.e., L2 loss. However, in the ---.cfg file, the output is a group of feature maps that is determined by the number of classes and the number of anchors.

So, the solution is that "First, defining a network that is same with the model in the ---.cfg file. Then, loading the learned parameters to this defined network and using 'model.save('model.h5')' to save this model. Final, convert it to any form".

This way has been verified on the task of converting a keras model (.h5) to a darknet model.

@vector-1127 Can you please share your code to highlight how this conversion is done

[krishnak]

File "D:yolov3keras-yolo3_phoneyolo3model.py", line 376, in yolo_loss
anchors[anchor_mask[l]], num_classes, input_shape, calc_loss=True)
TypeError: list indices must be integers or slices, not list

The reason for this error is anchors are not reshaped
anchors = np.array(anchors).reshape(-1, 2)

However we have no control over this when the model gets loaded from the file

[monchichi27]

I transform it sucessfully. It doesnot need to change save function, I use the below function to save weights and got 'trained_weights.h5'.
model.save_weights(log_dir + 'trained_weights.h5')

Then I use the below code:

from keras.layers import *
import os
import tensorflow as tf
from yolo3.model import yolo_body
def keras_to_tensorflow(keras_model, output_dir, model_name,out_prefix="output_", log_tensorboard=True):
if os.path.exists(output_dir) == False:
os.mkdir(output_dir)
out_nodes = []
for i in range(len(keras_model.outputs)):
out_nodes.append(out_prefix + str(i + 1))
tf.identity(keras_model.output[i], out_prefix + str(i + 1))
sess = K.get_session()
from tensorflow.python.framework import graph_util, graph_io
init_graph = sess.graph.as_graph_def()
main_graph = graph_util.convert_variables_to_constants(sess, init_graph, out_nodes)
graph_io.write_graph(main_graph, output_dir, name=model_name, as_text=False)
if log_tensorboard:
from tensorflow.python.tools import import_pb_to_tensorboard
import_pb_to_tensorboard.import_to_tensorboard(
os.path.join(output_dir, model_name),
output_dir)
keras_model = yolo_body(Input(shape=(None,None,3)), 3, 2) # my class_num=2
keras_model.load_weights("D:workspaceyolokeras-yolo3-masterlogs/000/trained_weights.h5")
output_dir = os.path.join(os.getcwd(),"checkpoint")
keras_to_tensorflow(keras_model,output_dir="D:workspaceyolokeras-yolo3-masterlogs/000",model_name="model.pb")
print("MODEL SAVED")

[Breezewrf]

With weight files, you can run model = yolo_body(Input(shape=(None, None, 3)), 3, num_classes) to create model structure, then model.load_weights('weights.h5') to load weights.

Your error 1, the model handles 20 classes, but you want it to handle 1. There is a conflict.
Error 2, yolo_head is in yolo3/model.py, so the error shouldn't happen.

great!
I have solved the problem
And visit my details at https://www.cnblogs.com/fragrant-breeze/p/12995593.html

[nengbunengzhengdianqi]

Excitedly, I have solved the problem.
The core problem is "the trained model that you want to save and the model defined in the ---.cfg file must be same".
Nacturally, during training, the model has only one output in this code, i.e., L2 loss. However, in the ---.cfg file, the output is a group of feature maps that is determined by the number of classes and the number of anchors.
So, the solution is that "First, defining a network that is same with the model in the ---.cfg file. Then, loading the learned parameters to this defined network and using 'model.save('model.h5')' to save this model. Final, convert it to any form".
This way has been verified on the task of converting a keras model (.h5) to a darknet model.

hello，I also got the issue when convert "yolo.h5" --> "yolo.pb". Follow your suggestions,

1. I have change model.save_weights() to model.save(),
2. but I don't understand the ---.cfg that above-mentionded , do you means the file yolov3.cfg that in the project?

The error information as follow:

  layer(input_tensors, **kwargs)
 File "/home/ql/anaconda3/envs/tf/lib/python3.5/site-packages/keras/engine/topology.py", line 619, in __call__
   output = self.call(inputs, **kwargs)
 File "/home/ql/anaconda3/envs/tf/lib/python3.5/site-packages/keras/layers/core.py", line 663, in call
   return self.function(inputs, **arguments)
 File "/home/ql/work/projects/keras-yolo3/yolo3/model.py", line 375, in yolo_loss
   grid, raw_pred, pred_xy, pred_wh = yolo_head(yolo_outputs[l],
NameError: name 'yolo_head' is not defined

Thanks for your help!

哥 解决了么 能教教我么 我也是在转化过程遇到这个Error

哥，这个问题你解决了吗