鏈接
https://www.52cv.net/?p=854
https://zhuanlan.zhihu.com/p/144530956
Abstract 概要
第一段
This paper proposes LPRNet - end-to-end method for Automatic License Plate Recognition without preliminary character segmentation.
本文提出了一種LPRNet -端到端車牌自動識別方法，該方法不需要進行初步的字符分割。
Our approach is inspired by recent breakthroughs in Deep Neural Networks, and works in real-time with recognition accuracy up to 95% for Chinese license plates: 3 ms/plate on nVIDIA GeForceTMGTX 1080 and 1.3 ms/plate on Intel R CoreTMi7-6700K CPU
我們的方法受到了深度神經網絡最近的突破的啟發，可以實時識別中國車牌，識別率高達95%:在nVIDIA GeForceTMGTX 1080上為3 ms/牌照，在Intel R CoreTMi7-6700K CPU上為1.3 ms/牌照
第二段
LPRNet consists of the lightweight Convolutional Neural Network, so it can be trained in end-to-end way.
LPRNet由輕量級的卷積神經網絡組成，可以端到端的方式進行訓練。
To the best of our knowledge, LPRNet is the first real-time License Plate Recognition system that does not use RNNs.
據我們所知，LPRNet是第一個不使用rnn的實時車牌識別系統。
As a result, the LPRNet algorithm may be used to create embedded solutions for LPR that feature high level accuracy even on challenging Chinese license plates.
因此，LPRNet算法可以用於創建嵌入式的LPR解決方案，即使在具有挑戰性的中國牌照上也具有高水平的准確性。
1.Introduction 介紹
第一段
Automatic License Plate Recognition is a challenging and important task which is used in traffic management, digital security surveillance, vehicle recognition, parking management of big cities.
車牌自動識別在大城市的交通管理、數字安全監控、車輛識別、停車管理等領域都有着重要的應用。
This task is a complex problem due to many factors which include but are not limited to: blurry images, poor lighting conditions, variability of license plates numbers (including special characters e.g. logograms for China, Japan), physical impact (deformations), weather conditions (see some examples in Fig. 1).
這個任務是一個復雜的問題，因為有很多因素，包括但不限於:模糊的圖像，糟糕的光照條件，可變性車牌號碼(包括特殊字符，如中國、日本的標識)、物理影響(變形)、天氣狀況(見圖1中的一些例子)。
第二段
license plates numbers (including special characters e.g. logograms for China, Japan), physical impact (deformations), weather conditions (see some examples in Fig. 1).
車牌號碼(包括特殊字符，如中國、日本的標識)、物理影響(變形)、天氣狀況(見圖1中的一些例子)。
第三段
This paper tackles the License Plate Recognition problem and introduces the LPRNet algorithm, which is designed to work without pre-segmentation and consequent recognition of characters.
本文針對車牌識別問題，介紹了LPRNet算法，該算法不需要進行字符的預分割和后續識別。
In the present paper, we do not consider License Plate Detection problem, however, for our particular case it can be done through LBP-cascade.
在本文中，我們不考慮車牌檢測問題，但在我們的特殊情況下，可以通過lbp級聯來實現。
第四段
LPRNet is based on Deep Convolutional Neural Network.
LPRNet是基於深度卷積神經網絡的。
Recent studies proved effectiveness and superiority
最近的研究證明了它的有效性和優越性
of Convolutional Neural Networks in many Computer Vision tasks such as image classification, object detection and semantic segmentation.
卷積神經網絡在許多計算機視覺任務中的應用，如圖像分類、目標檢測和語義分割。
However, running most of them on embedded devices still remains a challenging problem.
然而，在嵌入式設備上運行它們仍然是一個具有挑戰性的問題。
第五段
LPRNet is a very efficient neural network, which takes only 0.34 GFLops to make a single forward pass.
LPRNet是一種非常高效的神經網絡，它只需要0.34 GFLops就可以完成一次前進。
Also, our model is real-time on Intel Core i7-6700K SkyLake CPU with high accuracy on challenging Chinese License plates and can be trained end-to-end.
此外，我們的模型在英特爾酷睿i7-6700K SkyLake CPU上是實時的，對具有挑戰性的中國牌照具有很高的准確性，並且可以端到端訓練。
Moreover, LPRNet can be partially ported on FPGA, which can free up CPU power for other parts of the pipeline.
此外，LPRNet可以部分移植到FPGA上，這可以為流水線的其他部分釋放CPU功率。
Our main contributions can be summarized as follows:
我們的主要貢獻可以總結如下:
● LPRNet is a real-time framework for high-quality license plate recognition supporting template and character independent variable-length license plates, performing LPR without character pre-segmentation, trainable end-to-end from scratch for different national license plates.
● LPRNet是一個實時的高質量車牌識別框架，支持模板和字符獨立的變長車牌，對不同國家的車牌進行無字符預分割的LPR，從頭到尾可訓練。
● LPRNet is the first real-time approach that does not use Recurrent Neural Networks and is lightweight enough to run on variety of platforms, including embedded devices
● LPRNet是第一個不使用循環神經網絡的實時方法，它足夠輕量級，可以在各種平台上運行，包括嵌入式設備
● Application of LPRNet to real traffic surveillance video shows that our approach is robust enough to handle difficult cases, such as perspective and cameradependent distortions, hard lighting conditions, change of viewpoint, etc.
● LPRNet在實際交通監控視頻中的應用表明，該方法具有足夠的魯棒性，能夠處理諸如視角和攝像機依賴畸變、光照條件惡劣、視點變化等困難情況。
● The rest of the paper is organized as follows.
● 本文的其余部分組織如下。
● Section 2 describes the related work.
● 第2節描述相關的工作。
● In sec. 3 we review the details of the LPRNet model.
● 在第3節中，我們將回顧LPRNet模型的細節。
● Sec. 4 reports the results on Chinese License Plates and includes an ablation study of our algorithm.
● 第四部分報告了中國車牌的結果，並包括對我們算法的消融研究。
● We summarize and conclude our work in sec. 5.
● 我們在第5部分中總結和總結我們的工作。
2.相關工作
第一段
In the earlier works on general LP recognition, such as the pipeline consist of character segmentation and char acter classification stages：
在早期的一般LP識別工作中，如管道包括字符分割和字符分類階段
Character segmentation typically uses different handcrafted algorithms, combining projections, connectivity and contour based image components.
字符分割通常使用不同的手工算法，結合基於投影、連接性和輪廓的圖像組件。
It takes a binary image or intermediate representation as input so character segmentation quality is highly affected by the input image noise, low resolution, blur or deformations.
它以二值圖像或中間表示作為輸入，因此字符分割的質量受到輸入圖像噪聲、低分辨率、模糊或變形的嚴重影響。
Character classification typically utilizes one of the optical character recognition (OCR) methods - adopted for LP character set.
字符分類通常使用光學字符識別(OCR)方法之一-采用的LP字符集。
第二段
Since classification follows the character segmentation, end-to-end recognition quality depends heavily on the applied segmentation method.
由於分類是在字符分割的基礎上進行的，所以端到端識別的質量很大程度上取決於所采用的分割方法。
In order to solve the problem of character segmentation there were proposed endto-end Convolutional Neural Networks (CNNs) based solutions taking the whole LP image as input and producing the output character sequence.
為了解決字符分割問題，提出了一種基於端到端卷積神經網絡(CNNs)的字符分割方法，以整個LP圖像作為輸入，生成輸出字符序列。
第三段
The segmentation-free model in [2] is based on variable length sequence decoding driven by connectionist temporal classification (CTC) loss [3, 4].
[2]中的無分割模型是基於連接主義時序分類(CTC)損耗驅動的變長序列解碼[3,4]。
[ ] H. Li and C. Shen, “Reading Car License Plates Using Deep Convolutional Neural Networks and LSTMs,”
arXiv:1601.05610 [cs], Jan. 2016, arXiv: 1601.05610. 2,4
It uses hand-crafted features LBP built on a binarized image as CNN input to produce character classes probabilities.
它使用手工制作的特征LBP建立在二值化圖像作為CNN的輸入，以產生角色類別的概率。
Applied to all input image positions via the sliding window approach it makes the input sequence for the bi-directional Long-Short Term Memory (LSTM) [5] based decoder.
通過滑動窗口方法應用於所有的輸入圖像位置，使輸入序列用於雙向長短期記憶(LSTM)[5]解碼器。
Since the decoder output and target character sequence lengths are different, CTC loss is used for the pre-segmentation free end-to-end training.
由於解碼器輸出和目標字符序列長度不同，使用CTC損耗進行預分割自由端到端訓練。
第四段
The model in [6] mostly follows the approach described in [2] except that the sliding window method was replaced by CNN output spatial splitting to the RNN input sequence (”sliding window” over feature map instead of input).
[6]中的模型除了用CNN輸出空間分割到RNN輸入序列(feature map上的“滑動窗口”而不是輸入)來代替滑動窗口方法外，基本遵循了[2]中描述的方法。
[ ] T. K. Cheang, Y. S. Chong, and Y. H. Tay, “Segmentationfree Vehicle License Plate Recognition using ConvNetRNN,” arXiv:1701.06439 [cs], Jan. 2017, arXiv:
1701.06439. 2
第五段
In contrast [7] uses the CNN-based model for the whole LP image to produce the global LP embedding which is decoded to a 11-character-length sequence via 11 fully connected model heads.
而[7]對整個LP圖像使用基於cnn的模型產生全局LP嵌入，通過11個全連接的模型頭將全局LP嵌入解碼為11個字符長度的序列。
[ ] V. Jain, Z. Sasindran, A. Rajagopal, S. Biswas, H. S. Bharadwaj, and K. R. Ramakrishnan, “Deep Automatic License Plate Recognition System,” in Proceedings of the Tenth Indian Conference on Computer Vision, Graphics and Image Processing, ser. ICVGIP ’16. New York, NY, USA: ACM, 2016, pp. 6:1–6:8. 2
Each of the heads is trained to classify the i-th target string character (which is assumed to be padded to the predefined fixed length), so the whole recognition can be done in a single feed-forward pass.
每個頭部都經過訓練，對第i個目標字符串字符進行分類(假設該字符被填充到預定義的固定長度)，因此整個識別可以在一次前饋傳遞中完成。
It also utilizes the Spatial Transformer Network (STN) [8] to reduce the effect of input image deformations.
該算法還利用空間變壓器網絡(STN)[8]來降低輸入圖像變形的影響。
[ ] M. Jaderberg, K. Simonyan, A. Zisserman, and K. Kavukcuoglu, “Spatial Transformer Networks, arXiv:1506.02025 [cs], Jun. 2015, arXiv: 1506.02025. 2, 3
第56段
The algorithm in [9] makes an attempt to solve both license plate detection and license plate recognition problems by single Deep Neural Network.
[9]中的算法嘗試用單個深度神經網絡同時解決車牌檢測和車牌識別問題。
[ ] H. Li, P. Wang, and C. Shen, “Towards End-to-End Car License Plates Detection and Recognition with Deep Neural Networks,” ArXiv e-prints, Sep. 2017. 2
Recent work [10] tries to exploit synthetic data generation approach based on Generative Adversarial Networks [11] for data generation procedure to obtain large representative license plates dataset.
最近的工作[10]嘗試開發基於生成式對抗網絡[11]的合成數據生成方法，用於數據生成過程中獲取具有代表性的大型車牌數據集。
[ ] X. Wang, Z. Man, M. You, and C. Shen, “Adversarial Generation of Training Examples: Applications to Moving Vehicle License Plate Recognition,” ArXiv e-prints, Jul. 2017. 2
[ ] I. J. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio,
“Generative Adversarial Networks,” ArXiv e-prints, Jun. 2014. 2
第七段
In our approach, we avoided using hand-crafted features over a binarized image - instead we used raw RGB pixels as CNN input.
在我們的方法中，我們避免在二值化圖像上使用手工制作的特征，而是使用原始的RGB像素作為CNN輸入。
The LSTM-based sequence decoder working on outputs of a sliding window CNN was replaced with a fully convolutional model which output is interpreted as character probabilities sequence for CTC loss training and greedy or prefix search string inference.
將基於lstm的序列解碼器對滑動窗口CNN的輸出進行解碼，采用全卷積模型，將輸出解釋為字符概率序列，用於CTC損失訓練和貪婪或前綴搜索字符串推理。
For better performance the pre-decoder intermediate feature map was augmented by the global context embedding as described in [12].
為了獲得更好的性能，通過[12]中所述的全局上下文嵌入來增強預解碼器中間特征圖。
Also the backbone CNN model was reduced significantly using the low computation cost basic building block inspired by SqueezeNet Fire Blocks [13] and Inception Blocks of [14, 15, 16].
此外，利用受SqueezeNet Fire Blocks[13]和Inception Blocks[14,15,16]啟發的低計算成本的基本構建塊，大大減少了主干CNN模型。
Batch Normalization [17] and Dropout [18] techniques were used for regularization.
使用Batch Normalization[17]和Dropout[18]技術進行正則化。
第八段
LP image input size affects both the computational cost and the recognition quality [19], as a result there is a tradeoff between using high [6] or moderate [7, 2] resolution.
LP圖像輸入大小影響計算成本和識別質量[19]，因此需要在使用高[6]或中等[7,2]分辨率之間進行權衡。
[ ] S. Agarwal, D. Tran, L. Torresani, and H. Farid, “Deciphering Severely Degraded License Plates,” San Francisco, CA, 2017. 2
3 LPRNet
結構設計
第一段
In this section we describe our LPRNet network architecture design in detail.
在本節中，我們將詳細描述我們的LPRNet網絡架構設計。
In recent studies tend to use parts of the powerful classification networks such as VGG, ResNet or GoogLeNet as ‘backbone‘ for their tasks by applying transfer learning techniques.
在最近的研究中，通過應用遷移學習技術，傾向於使用部分強大的分類網絡，如VGG、ResNet或GoogLeNet作為他們任務的“骨干”。
However, this is not the best option for building fast and lightweight networks, so in our case we redesigned main ‘backbone‘ network applying recently discovered architecture tricks.
然而，這並不是構建快速和輕量級網絡的最佳選擇，所以在我們的案例中，我們應用最近發現的架構技巧重新設計了主要的“骨干”網絡。
第二段
The basic building block of our CNN model backbone (Table 2) was inspired by SqueezeNet Fire Blocks [13] and Inception Blocks of [14, 15, 16].
我們CNN模型主干的基本構建模塊(表2)的靈感來自於SqueezeNet Fire Blocks[13]和Inception Blocks[14,15,16]。
We also followed the research best practices and used Batch Normalization [17] and ReLU activation after each convolution operation.
我們也遵循研究的最佳實踐，並在每次卷積操作后使用Batch Normalization[17]和ReLU激活。
第三段
In a nutshell our design consists of:
簡而言之，我們的設計包括:
• location network with Spatial Transformer Layer [8]
•具有空間變壓器層[8]的定位網絡(optional)(可選)STN
• light-weight convolutional neural network (backbone)
•輕量級卷積神經網絡(骨干)
• per-position character classification head
•逐位字符分類頭
• character probabilities for further sequence decoding
•進一步序列解碼的字符概率
• post-filtering procedure
•后過濾過程
第四段
First, the input image is preprocessed by the Spatial Transformer Layer, as proposed in [8].
首先，按照[8]中提出的空間變換層對輸入圖像進行預處理。
This step is optional but allows to explore how one can transform the input image to have better characteristics for recognition.
這一步是可選的，但可以探索如何轉換輸入圖像以獲得更好的識別特征。
The original LocNet (see the Table 1) architecture was used to estimate optimal transformation parameters.
最初的LocNet(見表1)架構被用來估計最佳的轉換參數。
第五段
The backbone network architecture is described in Table 3.
骨干網的結構如表3所示。
The backbone takes a raw RGB image as input and calculates spatially distributed rich features.
主干采用原始RGB圖像作為輸入，並計算空間分布的豐富特征。
Wide convolution (with 1 × 13 kernel) utilizes the local character context instead of using LSTM-based RNN.
寬卷積(1 × 13核)利用局部字符上下文，而不是使用基於lstm的RNN。
The backbone subnetwork output can be interpreted as a sequence of character probabilities whose length corresponds to the input image pixel width.
主干網的子網絡輸出可以解釋為一個字符概率序列，其長度對應於輸入圖像的像素寬度。
Since the decoder output and the target character sequence lengths are of different length, we apply the method of CTC loss [20] - for segmentation-free end-to-end training.
由於解碼器輸出和目標字符序列長度不同，我們采用CTC損失[20]-的方法進行無分割端到端訓練。
CTC loss is a well-known approach for situations where input and output sequences are misaligned and have variable lengths.
當輸入和輸出序列不對齊且長度可變時，CTC損耗是一種眾所周知的方法。
Moreover, CTC provides an efficient way to go from probabilities at each time step to the probability of an output sequence.
此外，CTC提供了一種有效的方法，將每個時間步長的概率轉化為輸出序列的概率。
More detailed explanation about CTC loss can be found in .
關於CTC損失的詳細解釋請見。
第六段
To further improve performance, the pre-decoder intermediate feature map was augmented with the global context embedding as in [12].
為了進一步提高性能，在預解碼器中間特征圖中加入[12]中所示的全局上下文嵌入。
It is computed via a fully-connected layer over backbone output, tiled to the desired size and concatenated with backbone output.
它通過骨干網輸出的全連接層計算，平鋪到所需的大小，並與骨干網輸出連接。
In order to adjust the depth of feature map to the character class number additional 1 × 1 convolution is applied.
為了使特征圖的深度與字符類數相適應，采用了額外的1 × 1卷積。
For the decoding procedure at the inference stage we considered 2 options: greedy search and beam search.
對於推理階段的解碼過程，我們考慮了兩種選擇:貪婪搜索和波束搜索。
While greedy search takes the maximum of class probabilities in each position, beam search maximizes the total probability of the output sequence [3, 4].
貪婪搜索使每個位置的類概率最大，而波束搜索使輸出序列的總概率最大[3,4]。
For post-filtering we use a task-oriented language model implemented as a set of the target country LP templates.
對於后過濾，我們使用一組目標國家LP模板實現的面向任務的語言模型。
Note that post-filtering is applied together with Beam Search.
注意，后過濾是與波束搜索一起應用的。
The post-filtering procedure gets top-N most probable sequences found by beam search and returns the first one that matches the set of predefined templates which depends on country LP regulations.
后過濾程序根據波束搜索找到的最可能序列的top-N，根據國家LP規則返回與預定義模板集匹配的第一個序列。
訓練細節
All training experiments were done with the help of TensorFlow [21].
所有的訓練實驗都是在TensorFlow[21]的幫助下完成的。
We train our model with ’Adam’ optimizer using batch size of 32, initial learning rate 0.001 and gradient noise scale of 0.001.
我們使用“Adam”優化器訓練我們的模型，使用批量為32，初始學習率為0.001，梯度噪聲尺度為0.001。
We drop the learning rate once after every 100k iterations by a factor of 10 and train our network for 250k iterations in total.
在每100k次迭代之后，我們將學習速率降低10倍，並總共訓練我們的網絡250k次迭代。
In our experiments we use data augmentation by random affine transformations, e.g. rotation, scaling and shift.
在我們的實驗中，我們使用隨機仿射變換的數據增強，例如旋轉、縮放和移位。
It is worth mentioning, that application of LocNet from the beginning of training leads to degradation of results, because LocNet cannot get reasonable gradients from a recognizer which is typically too weak for the first few iterations.
值得一提的是，從訓練開始就應用LocNet會導致結果下降，因為LocNet無法從識別器獲得合理的梯度，而識別器在最初的幾個迭代中通常都太弱了。
So, in our experiments, we turn LocNet on only after 5k iterations.
所以，在我們的實驗中，我們只有在5000千米后才打開LocNet迭代。
All other hyper-parameters were chosen by crossvalidation over the target dataset.
所有其他超參數都是通過目標數據集的交叉驗證來選擇的。
結果&經驗
The LPRNet baseline network, from which we started our experiments with different architectures, was inspired by [2].
LPRNet基線網絡是受到[2]的啟發，我們從它開始用不同的架構進行實驗。
It’s mainly based on Inception blocks followed by a bidirectional LSTM (biLSTM) decoder and trained with CTC loss.
它主要基於Inception塊和雙向LSTM (biLSTM)解碼器，並使用CTC損耗進行訓練。
We first performed some experiments aimed at replacing biLSTM with biGRU cells, but we did not observe any clear benefits of using biGRU over biLSTM.
我們首先進行了一些旨在用biGRU細胞替換biLSTM的實驗，但是我們沒有觀察到使用biGRU比使用biLSTM有任何明顯的好處。
Then, we focused on eliminating of the complicated biLSTM decoder, because most modern embedded devices still do not have sufficient compute and memory to efficiently execute biLSTM.
然后，我們將重點放在消除復雜的biLSTM解碼器上，因為大多數現代嵌入式設備仍然沒有足夠的計算和內存來有效地執行biLSTM。
Importantly, our LSTM is applied to a spatial sequence rather than to a temporal one.
重要的是，我們的LSTM應用於空間序列而不是時間序列。
Thus all LSTM inpuuuts are known upfront both at the training stage as well as at the inference stage.
因此，在訓練階段和推理階段，所有的LSTM輸入都是預先知道的。
Therefore we believe that RNN can be replaced by spatial convolutions without a significant drop in accuracy.
因此，我們認為可以用空間卷積替代RNN，而精度不會有明顯下降。
The RNN-less model with some backbone modifications is referenced as LPRNet basic and it was described in details in sec. 3.
具有一些骨干修改的RNN-less模型被引用為LPRNet基礎模型，在第3節詳細描述了它。
To improve runtime performance we also modified LPRNet basic by using 2 × 2 strides for all pooling layers.
為了提高運行時性能，我們還修改了LPRNet basic，對所有池化層使用2 × 2的strides。
This modification (the LPRNet reduced model) reduces the size of intermediate feature maps and total inference computational cost significantly (see GFLOPs column of the Table 4).
這種修改(LPRNet縮減模型)大大減少了中間特征映射的大小和總推理計算成本(見表4的GFLOPs列)。
Ablation study消融實驗
It is of vital importance to conduct the ablation study to identify correlation between various enhancements and respective accuracy/performance improvements.
進行消融研究是至關重要的，以確定各種增強和各自的准確性/性能改進之間的相關性。
This helps other researchers adopt ideas from the paper and reuse most promising architecture approaches.
這有助於其他研究人員采納論文中的觀點，並重用最有前途的架構方法。
Table 5 shows a summary of architecture approaches and their impact on accuracy.
表5總結了架構方法及其對准確性的影響。
As one can see, the largest accuracy gain (36%) was achieved using the global context.
正如我們所看到的，使用全局上下文獲得了最大的准確性增益(36%)。
The data augmentation techniques also help to improve accuracy significantly (by 28.6%).
數據增強技術也有助於顯著提高准確度(提高28.6%)。
Without using data augmentation and the global context we could not train the model from scratch.
如果不使用數據擴充和全局上下文，我們就無法從頭開始訓練模型。
The STN-based alignment subnetwork provides noticeable improvement of 2.8-5.2%.
基於stn的對齊子網提供了2.8-5.2%的顯著改進。
Beam Search with postfiltering further improves recognition accuracy by 0.4- 0.6%.
帶有后濾波的波束搜索進一步提高了0.4- 0.6%的識別精度。
展望
In this work, we have shown that for License Plate Recognition one can utilize pretty small convolutional neural networks.
在這項工作中，我們已經證明了車牌識別可以利用相當小的卷積神經網絡。
LPRNet model was introduced, which can be used for challenging data, achieving up to 95% recognition accuracy.
引入LPRNet模型，可用於挑戰性數據，識別准確率高達95%。
Architecture details, its motivation and the ablation study was conducted.
進行了建築細節、動機和消融研究。
We showed that LPRNet can perform inference in realtime on a variety of hardware architectures including CPU, GPU and FPGA.
我們證明LPRNet可以在包括CPU、GPU和FPGA在內的各種硬件架構上實時執行推理。
We have no doubt that LPRNet could attain real-time performance even on more specialized embedded low-power devices.
我們毫不懷疑，即使在更專業的嵌入式低功耗設備上，LPRNet也可以實現實時性能。
The LPRNet can likely be compressed using modern pruning and quantization techniques, which would potentially help to reduce the computational complexity even further.
LPRNet可以使用現代剪枝和量化技術進行壓縮，這可能有助於進一步降低計算復雜度。
As a future direction of research, LPRNet work can be extended by merging CNN-based detection part into our algorithm, so that both detection and recognition tasks will be evaluated as a single network in order to outperform the LBP-based cascaded detector quality.
作為未來的研究方向，可以將基於cnn的檢測部分合並到我們的算法中，從而擴展LPRNet的工作，將檢測和識別任務作為一個單一的網絡進行評估，從而優於基於lbp的級聯檢測器的質量。