Automatic modeling method and device for object detection model
View Patent ↗An automatic modeling method for object detection model, the method includes: S 1, searching for an NAS network model according to a predetermined first neural network; S 2, training a first object detection model by using a training set, evaluating the current first object detection model by using a validation set after a number of training times reaches a first preset number of times, and outputting an evaluation result, wherein the first object detection model is a model obtained by detection part fusion of the NAS network model and a second object detection model; S 3, calculating a reward corresponding to the current first object detection model; S 4, adjusting parameters used for searching for the NAS network model according to the reward: repeating the steps S 1 -S 4 for a second preset number of times, and determining an optimal first object detection model according to the evaluation result. Further disclosed in the present invention is an automatic modeling device. The method and device provided by the present invention can search for different models according to different tasks, and improve an object detection effect.
1. An automatic modeling method for object detection model, comprising the following steps:
S 1 , searching for a Neural Architecture Search (NAS) network model according to a predetermined first neural network;
S 2 , training a first object detection model by using a training set, evaluating a current first object detection model by using a validation set after a number of training times reaches a first preset number of times, and outputting an evaluation result, wherein the first object detection model is a model obtained by detection part fusion of the NAS network model and a second object detection model;
S 3 , calculating an object detection task reward corresponding to the current first object detection model;
S 4 , adjusting parameters used for searching for the NAS network model according to the object detection task reward; and
repeating the steps S 1 -S 4 for a second preset number of times, and determining an optimal first object detection model according to the evaluation result;
wherein the NAS network model is a NAS network without a fully connected layer; the detection part fusion of the NAS network model and the second object detection model refers to: that a plurality of output ends of different scales of the NAS network model are each connected to a 1*1 convolutional layer, and an output of the 1*1 convolutional layer is used as an input of a detection part of the second object detection model.
2. The method according to claim 1 , wherein,
the predetermined first neural network is a recurrent neural network (RNN); and
the step S 1 comprises:
step S 11 , sampling through an RNN controller according to preconfigured hyperparameters, to obtain a subnetwork structure encoding; wherein said preconfigured hyperparameters include subnetwork hyperparameters and controller hyperparameters; the subnetwork hyperparameters include a number of layers of a subnetwork, a number of cell branches, parameters related to a learning rate of the subnetwork, and a number of output channels; the controller hyperparameters include parameters related to a learning rate of the RNN controller and optimizer configuration parameters; and
step S 12 : outputting the NAS network model by decoding the subnetwork structure encoding through a decoder.
3. The method according to claim 2 , wherein,
the adjusting parameters used for searching for the NAS network model according to the object detection task reward comprises:
feeding the object detection task reward back to the RNN controller, and updating trainable parameters of the RNN controller by θ←θ+γ∇ θ logπ θ (s t ,a t )·R t , wherein θ represents the trainable parameters of the RNN controller, γ∇ θ represents a back propagation gradient of the trainable parameter, γ represents a preset parameter, logπ θ (s t ,a t ) is a cross entropy loss corresponding to the NAS network model searched for when step S 1 is performed for a t-th time, and R t represents a reward value corresponding to the NAS network model searched for when step S 1 is performed for the t-th time.
4. The method according to claim 1 , wherein after determining the optimal first object detection model according to the evaluation result, the method comprises:
training the first object detection model with optimal evaluation result by using a training set, and testing the trained first object detection model with optimal evaluation result on a test set.
5. The method according to claim 1 , wherein,
the second object detection model includes you only look once (Yolo).
6. The method according to claim 1 , wherein,
the evaluating a current first object detection model by using a validation set comprises:
performing Mean Average Precision (MAP) evaluation on the current first object detection model by using the validation set;
the determining an optimal first object detection model according to the evaluation result comprises:
determining a first object detection model corresponding to a maximum MAP within the second preset number of times as the optimal first object detection model.
7. The method according to claim 1 , wherein the calculating an object detection task reward corresponding to the current first object detection model comprises:
reward
=
tan
(
(
α
1
L
cla
+
(
1
-
a
)
1
L
reg
-
baseline
)
*
π
2
)
,
where
(
α
1
L
cla
+
(
1
-
a
)
1
L
reg
-
baseline
)
∈
(
-
1
,
1
)
wherein L cla is classification loss, L reg is regression loss, and the baseline is a preset numerical value or formula.
8. An automatic modeling device for object detection model, comprising a memory and a processor, wherein,
the memory is configured for storing a program for automatic modeling of the object detection model; and
the processor is configured for reading and performing the program for automatic modeling of the object detection model, and performing the following operations:
S 1 , searching for a Neural Architecture Search (NAS) network model according to a predetermined first neural network;
S 2 , training a first object detection model by using a training set, evaluating a current first object detection model by using a validation set after a number of training times reaches a first preset number of times, and outputting an evaluation result, wherein the first object detection model is a model obtained by detection part fusion of the NAS network model and a second object detection model;
S 3 , calculating an object detection task reward corresponding to the current first object detection model;
S 4 , adjusting parameters used for searching for the NAS network model according to the object detection task reward; and
repeating the steps S 1 -S 4 for a second preset number of times, and determining an optimal first object detection model according to the evaluation result;
wherein the NAS network model is a NAS network without a fully connected layer; the detection part fusion of the NAS network model and the second object detection model refers to: that a plurality of output ends of different scales of the NAS network model are each connected to a 1*1 convolutional layer, and an output of the 1*1 convolutional layer is used as an input of a detection part of the second object detection model.
9. The device according to claim 8 , wherein,
the predetermined first neural network is a recurrent neural network (RNN); and
the step S 1 comprises:
step S 11 , sampling through an RNN controller according to preconfigured hyperparameters, to obtain a subnetwork structure encoding; wherein said preconfigured hyperparameters include subnetwork hyperparameters and controller hyperparameters; the subnetwork hyperparameters include a number of layers of a subnetwork, a number of cell branches, parameters related to a learning rate of the subnetwork, and a number of output channels; the controller hyperparameters include parameters related to a learning rate of the RNN controller and optimizer configuration parameters; and
step S 12 : outputting the NAS network model by decoding the subnetwork structure encoding through a decoder.
10. The device according to claim 9 , wherein,
the adjusting parameters used for searching for the NAS network model according to the object detection task reward comprises:
feeding the object detection task reward back to the RNN controller, and
updating trainable parameters of the RNN controller by θ←θ+γ∇ θ logπ θ (s t ,a t )·R t , wherein θ represents the trainable parameters of the RNN controller, γ∇ θ represents a back propagation gradient of the trainable parameter, γ represents a preset parameter, logπ θ (s t ,a t ) is a cross entropy loss corresponding to the NAS network model searched for when step S 1 is performed for a t-th time, and R t represents a reward value corresponding to the NAS network model searched for when step S 1 is performed for the t-th time.
11. The device according to claim 8 , wherein the processor is configured for reading and performing the program for automatic modeling of the object detection model, and performing the following operations:
after determining the optimal first object detection model according to the evaluation result, training the first object detection model with optimal evaluation result by using a training set, and testing the trained first object detection model with optimal evaluation result on a test set.
12. The device according to claim 8 , wherein,
the second object detection model includes you only look once (Yolo).
13. The device according to claim 8 , wherein,
the evaluating a current first object detection model by using a validation set comprises:
performing Mean Average Precision (MAP) evaluation on the current first object detection model by using the validation set;
the determining an optimal first object detection model according to the evaluation result comprises:
determining a first object detection model corresponding to a maximum MAP within the second preset number of times as the optimal first object detection model.
14. The device according to claim 8 , wherein the calculating an object detection task reward corresponding to the current first object detection model comprises:
reward
=
tan
(
(
α
1
L
cla
+
(
1
-
a
)
1
L
reg
-
baseline
)
*
π
2
)
,
where
(
α
1
L
cla
+
(
1
-
a
)
1
L
reg
-
baseline
)
∈
(
-
1
,
1
)
wherein L cla is classification loss, L reg is regression loss, and the baseline is a preset numerical value or formula.