Scribd
Upload
13 views2 pages

Script

Traffic sign recognition is a difficult problem that is important for self-driving cars and driver safety. The document discusses the key topics of introducing sign recognition, the German Traffic Sign Recognition Benchmark dataset, literature on matrices, logistic regression, convolutional neural networks, and the proposed model architecture. Experimental results showed the model achieved 97.92% accuracy on training data and 87.92% on validation data, though not 100%.

Uploaded by

phungtienthanh2004
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
13 views2 pages

Script

Traffic sign recognition is a difficult problem that is important for self-driving cars and driver safety. The document discusses the key topics of introducing sign recognition, the German Traffic Sign Recognition Benchmark dataset, literature on matrices, logistic regression, convolutional neural networks, and the proposed model architecture. Experimental results showed the model achieved 97.92% accuracy on training data and 87.92% on validation data, though not 100%.

Uploaded by

phungtienthanh2004
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

Ladies and gentlemen, today I am pleased to dive into the fascinating field of Traffic Sign

Recognition, a challenge that has captivated the minds of computer scientists for over a decade.
As we know, automakers have made strides in developing solutions to enhance driver
convenience. However, the importance of Traffic Sign Recognition is not merely convenience, It
is the foundation for the development of self-driving cars, heralding the future of specialized
transportation.

This is today key topics and later we will learn more about them.

1. Intro
Sign recognition is a difficult problem for the world because it is related to traffic safety and has
puzzled scientists over the past decades. Sign recognition is necessary software to create self-
driving car models. Not only that, it also helps keep drivers safe.
2. dataset

The dataset we used here is German Traffic Sign Recognition Benchmark (GTSRB) because it
contains 43 types of traffic signs, training and testing images taken under real conditions, as
shown in the picture 3, a total of more than 50,000 images
3. literature review
In the field of Traffic Sign Recognition, a basic understanding of mathematics is very important.
The matrix plays a key role, defining vectors such as row vector 𝑥 and column vector 𝑥, whose
arrangement forms matrices such as 𝑋.

Unlike linear regression, logistic regression models binary outcomes. The logistic regression
equation, guided by the coefficients 𝑏0 and 𝑏1, becomes a powerful tool for this purpose.
Determined by the learning rate set by the user, Gradient Descent becomes the cornerstone of
our strategy in fine-tuning the Logistic Regression Model.

Last but not least, we enter the world of Convolutional Neural Networks .Proposr model
Now we will come to the proposal model
4. propose model
Now let's explore the complexity of our neural network architecture, an important factor in
pursuing our Traffic Sign Recognition challenge.

The above figure serves as our blueprint, outlining the architecture of our developed network,
based on the Convolutional Neural Network model. The above diagram includes input, data
processing and output. In the data processing section, small blocks are included for more
detailed processing.

The training process will be divided into two phases: forward and backward propagation. in
forward propagation we provide the predicted value 𝑦 based on the input 𝑥 and it allows the
model to calculate the error based on the loss function. In the backward propagation, we will
give data X and the model will calculate according to the above formula and produce results for
processing.
Put it all together
We can now have a big picture of how to train the model:

1) 𝐺𝑎𝑡ℎ𝑒𝑟 𝑡𝑟𝑎𝑖𝑛𝑖𝑛𝑔 𝑑𝑎𝑡𝑎𝑠𝑒𝑡 (𝑋train , 𝑌train ) 𝑎𝑛𝑑 𝑡𝑒𝑠𝑡 𝑑𝑎𝑡𝑎𝑠𝑒𝑡 (𝑋test , 𝑌test
).
2) 𝐼𝑛𝑖𝑡𝑖𝑎𝑙𝑖𝑧𝑒 𝑚𝑜𝑑𝑒𝑙 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟𝑠 𝑤 𝑎𝑛𝑑 𝑏.
3) 𝐷𝑒𝑓𝑖𝑛𝑒 𝑙𝑒𝑎𝑟𝑛𝑖𝑛𝑔 𝑟𝑎𝑡𝑒 𝛼, 𝑎𝑛𝑑 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑖𝑛𝑖𝑛𝑔 𝑖𝑡𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑠.
4) 𝐹𝑜𝑟 𝑒𝑎𝑐ℎ_𝑡𝑟𝑎𝑖𝑛𝑖𝑛𝑔_𝑖𝑡𝑒𝑟𝑎𝑡𝑖𝑜𝑛 𝑑𝑜
 𝑎𝑝𝑝𝑙𝑦 𝑓𝑜𝑟𝑤𝑎𝑟𝑑 𝑝𝑟𝑜𝑝𝑎𝑔𝑎𝑡𝑖𝑜𝑛 𝑤𝑖𝑡ℎ(𝑋train , 𝑌train ), 𝑤 𝑎𝑛𝑑 𝑏
 𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒 𝑐𝑜𝑠𝑡 𝑓𝑢𝑛𝑡𝑖𝑜𝑛 𝐽
 𝑎𝑝𝑝𝑙𝑦 𝑏𝑎𝑐𝑘𝑤𝑎𝑟𝑑 𝑝𝑟𝑜𝑝𝑎𝑔𝑎𝑡𝑖𝑜𝑛 𝑡𝑜 𝑎𝑑𝑗𝑢𝑠𝑡 𝑤 𝑎𝑛𝑑 𝑏
5) 𝑉𝑒𝑟𝑖𝑓𝑦 𝑡ℎ𝑒 𝑎𝑐𝑐𝑢𝑟𝑎𝑐𝑦 𝑜𝑓 𝑡ℎ𝑒 𝑚𝑜𝑑𝑒𝑙 𝑢𝑠𝑖𝑛𝑔 (𝑋test , 𝑌test )

5. Experiment result
This is the result of the sign recognition program that we tested and it produced unexpected
results with an accuracy of 97.92% in the 9th epoch of the training data set and 87, 92% of the
data set is valid.

However, the accuracy doesn’t reach 100%. Figure on the left shows the images of traffic signs
that were identify incorrectly. As it is seen in the figure, the quality of input images strongly
influences on the classification rate.

6. Conclusion
From the experimental results above, it can be said that the program we wrote worked quite
well and gave almost completely accurate results. This model could provide information related
to driving roads and improves safety mechanisms.
7. Q&A

You might also like