Control of Traffic Signals by AI based Image Processing

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 10 | Oct 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 212
Control of Traffic Signals by AI based Image Processing
Mareddy Pragnya Bharathi(1), P. Venkat Narayana(2)
Dept. of Electrical and Electronics Engineering, JNTU College of Engineering, Hyderabad, India
----------------------------------------------------------------------***----------------------------------------------------------------------
Abstract— The blockage of traffic in urban cities is
one of the major issues in the cities with the
population growth and increase in automobiles. In
the present-day Traffic system, wastage of fuel
consumption in non-peak hours, as the ongoing
traffic signal gives green light for longer duration,
even there are few vehicles on the road, and in some
cases, the user must wait as for signal to turn green
while there are no other vehicles in the junction's
other lanes, which adds to the driver's delay,
increases fuel consumption, and raises
transportation expenses, and increase air pollution.
Regulating traffic flow in peak hours for vehicles to
pass is the crucial factor. The ongoing traffic
management systems that are in place are generally
static, means that they do not adjust according to
needs of the traffic flow. The suggested solution
aims to provide a computer vision-based traffic
controller that can adjust to the flow of traffic. By
identifying the passing vehicles at the signal and
adjusting green signal period accordingly, it
employs a live feed from CCTV footage at traffic
intersections to evaluate the traffic density in real-
time. The automobiles are categorized as cars,
motorbike, bus, truck, or tricycle to determine the
green signal time with greater accuracy. For
detecting the objects, techniques like YOLO
algorithm are used and also counts number of
automobiles for each direction. The timers of these
traffic signals are set according to density of vehicle
in each lane and hence the system becomes
adaptive. This helps for clearing the traffic and
optimizing the signal as green at a faster rate
compared to present system, thus reduces the
unnecessary delays, blockage, and waiting time.
This reduces consumption of fuel and pollution.
Pygame is used for developing simulation.
Keywords—Traffic management, Artificial
Intelligence, Object detection with YOLO algorithm,
Computer Vision.
I. INTRODUCTION
With an increase in traffic, many road networks in
urban areas experience issues with road capacity and
associated service levels. Numerous traffic-related
problems arise at the intersection control systems as a
result of constant signal timers. They continue with the
exact phase sequence as well as its duration. The need
for cutting-edge control solutions will be identified in the
field of smart transportation systems as the need for road
capacity rises.
From the analysis of Traffic in Hyderabad from “The
Times of India” The problem, according to officials of the
police department and Road Transport Authority (RTA),
has been triggered by a huge jump in private vehicular
population in the city during the pandemic.
An estimated 10 lakh new vehicles were registered
within the city, including just 4.5 lakh vehicles in the
Cyberabad limits alone, resulting in an explosion within
the population of vehicles which together with the
present ones are criss-crossing the town, as per RTA
authorities
There are conventional techniques for control that are
getting used currently:
1)Controlling with static timers at traffic signals: In these
fixed timers are loaded into the program. Even when
there's no traffic on the lane the green signal is going to
be present supported the timer set within the program.
This causes delay for people waiting at junction, and also
increases fuel consumption, transportation cost and
increase pollution.
2)Controlling with Manpower: They are managed by a
Traffic Police standing at junction controlling the vehicles
by hand gestures.
This method requires lot of manpower, when traffic is
heavy and increasing day by day.
3)Detecting the Traffic with air-sensors: These devices
detect the presence of pollution in environment. If
pollution reaches particular level, depending on the
extent of pollution, traffic density is detected using a
sensor.
Since the vehicles on opposite lane also produces
pollution, this is drawback for calculating the traffic
density.
4)Sensors on the road: Adding some loop detectors or
proximity sensors to the road is another advanced
technique. This sensor transmits data on the movement
of traffic. Collected sensor data is used to manage various
traffic signals.

This contains a drawback of precision and scope of
high-quality information i.e., usually supported high-
cost technologies; hence a tight budget will cut back on
the number of facilities. Additionally, a lot of sensors are
typically needed to provide complete coverage across a
facility network due to the limited optimal range for the
most sensors.
Proposed system: The suggested system in this paper
is meant to detect the volume of traffic and timers are
set for every lane based on the traffic density calculated,
this density may easily be adapted to current state of
the traffic. Here, calculating traffic density can be done
by taking live video at the junction, and by identifying
the number of automobiles for individual lane
approaching the junction and adjusting the duration of
the green light. To determine a precise estimation of a
green traffic time, the automobiles are identified as
being a car, motorbike, bus, truck, or tricycle. It uses
YOLO, so as to identify the number of vehicles and so set
the timer of particular light in step with vehicle amount
within the appropriate direction.
This reduces unneeded delays, overcrowding, and
waiting times, which successively reduces the amount of
fuel used and pollutants produced. Additionally, traffic
is moved much more quickly than it would be under a
static system.
II. LITERATURE REVIEW
A traffic management system supported image
processing utilizing MATLAB code that modifies the
timing of the green, yellow, red light depending on the
volume of traffic. One Arduino UNO is used to control
the green and yellow lights, and the other is used to
control the red light and this is often endless process.[6]
An adaptive stoplight controller supported symbolic
rationale for enhancing traffic flow at a remote
crossroads. Using the data gathered from traffic
detectors, a set of fuzzy rules has been created (queue
length, arrival flow, and exit flow), evaluates the time
for the lane. Two components make up the suggested
fuzzy system: one with the primary driveway (which
will receive the next volume of vehicles), and one for the
second driveway (with a lower volume of vehicles).
Different scenarios with varied traffic demands are used
to compare proposed controller to a set signal
programmed in order to verify the performance of the
proposed decision rules.[9]
Before being submitted to the databases, the live
video feed is edited. A C++-based algorithms is then
used to generate the results. The techniques of hard
coding and dynamic coding are contrasted, and the
dynamic approach demonstrated a 35% improvement
for controlling the signal timers.[3]
PIC microcontroller that implements dynamic time
slots with various levels and measures traffic volume
using Infrared sensor. Additionally, conveyable
controller is designed to resolve traffic density and set
the timers for overcrowded roads.[4]
A camera installed with light accustomed captures the
live road images. Then the acquired pictures are loaded
into digital image processor to seek out the traffic density
on the road then the traffic lights are managed. By
avoiding time lost on empty roads, the suggested strategy
makes optimal use of time and fuel resources.[5]
Using image and video processing, actual traffic
volume may be calculated using live video stream from
the webcams at traffic intersections. It also works upon
that algorithm for adjusting traffic signals in accordance
with traffic volume on the road, intending to lessen
congestion on roads that can reduce the number of
accidents and fuel consumption will be reduced over
time. Additionally, it'll still offer important information
that can support analysis and planning for upcoming road
projects. To further reduce traffic snarls and improve
traffic flow, moreover signal lights are frequently synced
with one another.[6]
Photos are captured by the camera using a
Microcontroller based system, which then processes the
photographs in MATLAB. Vehicle density is determined
after the images are converted to criterion images by
reducing saturation and colors. Through the USB and
preconfigured simulation packages, MATLAB and
Arduino are coupled. The Arduino determines the length
of each lane's green light based on traffic volume and
density. But there is a problem with this approach. The
vehicles frequently collide one another and it's difficult to
induce correct number of vehicles on the road.
Additionally, since diverse things were also turned to
white and black, they interfered with detection and is not
possible to distinguish between common things like
poles, trees, and billboards with vehicles.[7]
Algorithm for vector machines together using image
processing techniques. in real-time video, the method is
used because images in short frames are recorded.
Prior to using SVM, OpenCV is used to complete the
image processing and convert the images to grayscale.
Using this technique, red light violations and traffic
density are both detected.[8]
A traffic density- and image-processing-based
adaptive light timing control. This method uses a high-
resolution image sensor, MATLAB, an Arduino signal
timing, and transmission based on UART principles. But,
neither the permitted ambulances nor accidents at the
intersection are prioritized by this system.[9]

III. SUGGESTIVE SYSTEM
1) Overview of Proposed System:
In Proposed system live video is taken at traffic
intersection as a source for current traffic density
calculation. Here, density calculation utilizes image
processing with python object detection is carried out
utilizing YOLO Algorithm. This image is given to the
vehicle identification system, which employs YOLO, as
seen in Fig. 1. The YOLO method counts the number of
each sort of vehicle, including cars, bicycles, buses, and
trucks, to evaluate the traffic density. This density, along
with a few other variables, is used by the signal
changing algorithm to fix the signal time as green for
each lane. Accordingly, the signal as red are modified. In
order to prevent a particular lane from going without
traffic, the green signal timer is limited to a
minimum and maximum value. Additionally, simulation
is created to show the precision and contrast it with the
current static system.
2. Module for detecting vehicle:
The suggested system leverages "You Only
Look Once" (YOLO) for object classification, which offers
the requisite precision and processing speed. Vehicle
detection was taught utilizing a customized YOLO
model, which can identify a number of vehicles
including automobiles, bikes, large vehicles (such as
trucks and buses), and motorbikes
Fig. 1. Proposed System Model
The YOLOv3 algorithm first separates a picture into a
grid. a grid cell each predicts some number of boundary
boxes (sometimes stated as anchor boxes) around objects
that score highly with the aforementioned predefined
classes.
Yolo takes an input image initially:
The framework creates grids from input image (say a 3 X
3 grid):
Each grid employs image localization and
classification. The object's bounding boxes and the
associated class probabilities are then predicted by YOLO.
Suppose we've got divided the image into a 3 x 3 grid and
have created a total of three categories into which we
want the object to be categorised. Consider that the
categories are, respectively, Motorcycle, Car, and
Pedestrian. So, for every grid, the label y is going to be an
eight-dimensional vector:
Here,
 pc determines whether or not an object is included in
the grid (it is the probability)
 by, bx, bw, and bh specifies the bounding box if any
object is present

 c1, c2, and c3 stand in for the classes. In the event
that the object is an automobile, then c2 will be 1
and both c1, c3 will be 0, and the process continues.
Suppose we choose a first grid as from example
above:
Because this grid is empty, pc will be 0 and the label y
of grid be:
The grid is empty, therefore the bx, by, bh, bw, c1, c2,
and c3 contain. Take another grid where an automobile
is present (c2 = 1)
YOLO determines whether an actual object is
present within the grid. If there are two objects (two
automobiles) in the image YOLO takes midpoint among
those two items and assign those objects to the grid that
contains the centroid of those objects. Then car's center-
left grid's y label reads
Since we have an object in this grid, pc is enough to
1. then by, bx, bw, and bh will be calculated in relation
to the specific grid cell that we are working with. Cars
being of the second class, c2 = 1 & c3 and c1 = 0,
respectively. We will therefore obtain an eight
dimension output vector for each of the nine grids. The
output will be three by three by eight in size. So now we
have a target vector and an input image. Our model will
be trained using the example above (input image: 100 X
100 X 3, output: 3 X 3 X 8):
Encoding bounding boxes:
In relation to the grid cell that we are working with,
the by, bx, bw, and bh are calculated. Take a look at the
right-center grid, which has an automobile in it:

Therefore, just this grid will be used to calculate by,
bx, bw, and bh. This grid's label y will be:
pc = 1 Since a car is present in this grid and there's an
object, so c2 = 1. The coordinates for any or all of the
grids (by, bx, bw, and bh) in YOLO are:
The by, bx are the x and y coordinates of the
midpoint of the article with relevancy this grid. during
this case:
The ratio of the height of the bounding box (the red box
in the example above) to the height of the associated
grid cell is known as bh, and in our case, it is
approximately 0.9. So, bh = 0.9. BW is the proportion of
the bounding box's width to the width of a grid cell. So,
bw = 0.5 (approximately). This grid's y label will be:
The midpoint will always be inside the grid, hence the
by and bx will always fall between 0 and 1. While bh &
bw are frequently equal to 1, just in instance the
bounding box's dimensions are equal to the grid's
dimensions.
Implementation:
Step-1: Required libraries are imported
Step-2: Create a method to filter individual boxes
depending on their criteria and chances
Step-3: Develop an algorithm to determine the IOU
among two boxes
Step-4: Implement a Non-Max Reduction function
Step-5: Estimate the bounding box coordinates after
creating a random volume with the size (19,19,5,85).
Step-6: Finally, we will build a function that will
accept the CNN outputs as input and produce the
suppressed boxes.
Step-7: Make a prediction for a random space using
the yolo eval utility
Step-8: Test a previously trained YOLO algorithm on
recent photos.
Step-9: Create a method to anticipate the bounding
boxes, then save the photos with them.
Step-10: Using the estimate function, read a picture
and estimation
Step-11: Plot the estimations
Here videos of 4 lanes at a junction are taken and
detected vehicles for 5 sec each on each side of the lane
for calculation of vehicle density This is a picture of
detected vehicles in the lane's right side
Detected vehicles on right lane = 57
Similarly, front lane detected vehicles =16

Left lane detected vehicles= 28
Back lane detected vehicles= 12
Considering the detected vehicles, Vehicle Traffic flow
and vehicle can be calculated:
1)Vehicle Traffic Flow
The vehicle traffic flow on a road can be
expressed q=n*3600/t
where,
q = vehicle flow (no. vehicles per hour)
n = no. vehicles passing in t seconds
t = time for passing vehicles (s)
2) Vehicle density:
K=(n*1000)/l
Where,
k = vehicles per km
n = no. of vehicles occupying a length l on the road
l = the length of road occupied by vehicles (m)
Considering traffic density calculated, the timer will be
determined by the density on the road.
1) Module of signal switching
The vehicle detecting module's traffic density
data is used by the Signal switching algorithm to set
green traffic signal timer and update red traffic signal
timers of all other signals. Additionally, it cycles back and
forth among the signals in accordance with timers.
The detection module's vehicle detection
information, as described in the preceding section, serves
as the algorithm's input. This one is frequently in JSON
format, also with accuracy and dimensions as the values
and the label of the object detected as the key. After this
data has been analyzed, the total number of automobiles
in each class is determined.
Following that, the signal's green time is estimated and
assigned, and also the signal’s red time of all other signals
are altered correspondingly. The method can be adjusted
for any multiple of signals at a junction, either up or
down.
The following elements are taken into account when
creating the algorithm:
a) The time interval of algorithm to estimate traffic
density and the length of green signal light, this gives
information of image to be acquired at particular time
b) Count of lanes
c) Total number of each type of vehicle, including
automobiles, trucks, motorbikes, etc.,
d) Using the aforementioned variables, traffic density
was determined
e) Each car has lag at startup, and also the non-linear
rise in lag faced by the vehicles in the rear results in
additional time.

f) The avg speed of every class of car whenever the
green light turns on, or the average amount of time
each type of vehicle needs to cross the intersection.
g) The maximum and minimum cut-off time for the
duration of a green signal, to stop blockage of traffic
The standard time for the principal signal of an
initial cycle is ready when the algorithm is run for
the first time, and as a result, the algorithm sets the
times for all other indicators of the initial cycle and all
subsequent signals. Every direction's vehicle detection
is handled by a different thread, while the signal's
timer is handled by the main thread. The detecting
threads take a picture of the following direction
whenever the green traffic timer of the present signal
(or the red-light timer of the next green light) reaches 5
seconds. The timer for the succeeding green signal is
then set when the result has been parsed. All of this
occurs within background as the majority of threads
continue to count down the remaining seconds on the
green signal timer. This makes it possible for the timer
to be assigned seamlessly, avoiding any latency. The
succeeding signal turns green for number of seconds
set by the algorithm whenever the green timer of this
signal reaches zero.
When the signal that will turn into green next is
5 seconds away, the picture is taken. This gives the
system a full 10 seconds to analyze the image, count
the number of vehicles in every class that are visible
there, determine the green traffic time, and set the
duration of this signal as well as the red signal period
of the next signal correspondingly. The normal speeds
of cars at start and their acceleration times have been
used, based on which an estimation of the typical time
each class of vehicle takes to cross a junction was
established, to get the ideal green signal time
supported the number of vehicles of each class at a
signal. Then, the green traffic time is determined using
(1)
∑
Here:
• GST means Green Signal Time
• NoofVehiclesofClass means number of cars of each
vehicle class on the signal detected by the module of
vehicle detection.
• AverageTimeofClass means average time for vehicles
of that class to cross the intersection.
• NoofLanes means Number of lanes in the junction.
The average-time it takes for each class of vehicles to
cross an intersection is set according to the placement.
i.e., By region, by city, by site, or perhaps by
intersection-wise, assisted by the property of the
intersection to facilitate traffic management. For this
purpose, data from each transport company are often
evaluated.
Instead of always pointing in the direction closest to
the start, the signal periodically changes. This
frequently occurs when the present system is
coordinated, so people don't have to reroute or cause
confusion and traffic lights turn green one at a time in a
set pattern. Additionally, because of the existing system,
traffic lights are in the same order and yellow traffic
lights are regarded the same way.
Order of signal: Red to Green to Yellow to Red
4) Module of simulation
To model actual traffic, Pygame was used to create
simulation from scratch. It helps with system
visualisation and contrast with the current static
system. A 4-way junction with 4 traffic lights is present
there. Each signal has a timer on its top that indicates
how long it will take for the light to change from green--
yellow, red--green, or from yellow--red. The number of
vehicles that have passed through the intersection is
also shown next to each light. Automobiles, bicycles,
buses, lorries, and rickshaws are all readily available. A
few of the automobiles in the rightmost lane
communicate cross the junction to make the simulation
seem more realistic. When a vehicle is generated, the
ability to rotate or not is set up using random data. It
also has a timer that shows how much time has passed
since the simulation started. A snapshot of the
simulation's final result is shown in Fig. 2.
Fig 2: simulation output
Lane-wise Vehicle Counts
Lane 1: 92
Lane 2: 105

Lane 3: 26
Lane 4: 27
Total vehicles passed: 250
Total time passed: 300
No. of automobiles passed per unit time:
0.833333334
A Python module set called Pygame is cross-platform
and made for creating video games. This provides music
and graphics libraries created specifically for use with
Python artificial language. On top of the superb SDL
library, Pygame adds capabilities. This enables Python
programmers to develop multimedia applications and
fully functional games. Pygame could also be used on
almost any platform and package and is fairly portable.
In the below image 4 sides of the junction is taken
and vehicle detection is done in each lane, according to
the vehicle detection count, timer is set green for more
time on the side of lane with more vehicles than the
other side with less vehicles.
Figure 3: 4 lanes output
Lane-wise Vehicle Counts
Lane 1 : 57
Lane 2 : 16
Lane 3 : 28
Lane 4 : 12
Fig4: Results of simulation
IV) ANALYSIS AND OUTCOMES
A. Evaluation of auto Recognition Module
Applying various test photos comprising varied
numbers of cars, the vehicle detection-module was put
to the test, and as a result, it was determined that
between 75% and 80% of time, detection accuracy was
accurate. In above Fig. 4, test result is displayed. This is
frequently acceptable but not ideal. The absence of an
accurate dataset is the primary cause of low accuracy.
To build on this, model is frequently trained using
actual traffic camera footage, increasing the system's
accuracy.
B. Assessment of the suggested adaptive system
No matter of distribution, the suggested
system always outperforms the static system. The
improve in efficiency is based on how unevenly traffic is
distributed among the lanes. The performance improves
as the traffic distribution becomes more skewed.
The suggested approach only marginally outperforms
the current system when traffic is evenly or near to
evenly distributed throughout the 4 lanes. Here,
performance has improved by around 9%.
The suggested system outperforms current
method by a large margin when the traffic distribution
is considerably skewed. Here, performance has
improved by roughly 22%. This is often the nature of
traffic dispersion that occurs in real-world settings.
The suggested system offers a vast performance gain
than the exsisting system when the dispersion of traffic
was significantly skewed. where the red and mete are
separated by an excessive space and the line descends
quickly. Here, performance has improved by roughly
36%.
The simulations are conducted for 5 min for
every distribution under identical simulation
conditions, including the dispersion of traffic, vehicle

speeds, that vehicles would turn, the distance between
vehicles, etc. It was identifiably demonstrated that the
suggested system performed 23% better on average than
the existing system with set times. This also suggests a
decrease in the length of time that green light stays idle
because of the vehicles' waiting period
V. EPILOGUE
In conclude, the suggested system provides the
signal as green for the lane with more traffic for a
extended period of time than the lane with less traffic by
setting the green signal time adaptive with the traffic
density at the junction. This can lessen unneeded delays,
traffic, and waiting times, which will ultimately result in
less fuel usage and pollution.
In terms of the quantity of vehicles passing the
junction, the system appears to perform around 23%
better than the current method, which might be a
considerable improvement. This method will be
enhanced to achieve even greater results with further
calibration utilizing real-world CCTV data for the model
training.
REFERENCES
[1] M. M. Gandhi, D. S. Solanki, R. S. Daptardar and N. S.
Baloorkar, "Smart Control of Traffic Light Using
Artificial Intelligence," 2020 5th IEEE International
Conference on Recent Advances and Innovations in
Engineering (ICRAIE), 2020, pp. 1-6, doi:
10.1109/ICRAIE51050.2020.9358334.
[2] https://www.analyticsvidhya.com/blog/2018/12/pr
actical-guide-object-detection-yolo-framewor-
python/
[3] https://www.researchgate.net/publication/2241066
88_Design_of_Intelligent_Traffic_Light_Controller_Usi
ng_Embedded_System
[4] A. Vogel, I. Oremović, R. Šimić and E. Ivanjko,
"Improving Traffic Light Control by Means of Fuzzy
Logic," 2018 International Symposium ELMAR, 2018,
pp. 51-56, doi: 10.23919/ELMAR.2018.8534692.
[5] J. Hosur, R. Rashmi and M. Dakshayini, "Smart Traffic
light control in the junction using Raspberry PI,"
2019 3rd International Conference on Computing
Methodologies and Communication (ICCMC), 2019,
pp. 153-156, doi: 10.1109/ICCMC.2019.8819695.
[6] Khushi, "Smart Control of Traffic Light System using
Image Processing," 2017 International Conference on
Current Trends in Computer, Electrical, Electronics
and Communication (CTCEEC), 2017, pp. 99-103, doi:
10.1109/CTCEEC.2017.8454966.
[7] .https://timesofindia.indiatimes.com/city/hyderab
ad/only-40-tech-firms-open-yet-traffic-jams-
plague-it corridor/articleshow/92760259
[8] A. A. Zaid, Y. Suhweil and M. A. Yaman, "Smart
controlling for traffic light time," 2017 IEEE Jordan
Conference on Applied Electrical Engineering and
Computing Technologies (AEECT), Aqaba, 2017, pp.
1-5, doi: 10.1109/AEECT.2017.8257768.
[9] Renjith Soman "Traffic Light Control and Violation
Detection Using Image Processing”.” IOSR Journal of
Engineering (IOSRJEN), vol. 08, no. 4, 2018, pp. 23-
27
[10] Siddharth Srivastava, Subhadeep Chakraborty, Raj
Kamal, Rahil, Minocha, “Adaptive traffic light timer
controller” , IIT KANPUR, NERD MAGAZINE.
[11] A. Kanungo, A. Sharma and C. Singla, "Smart traffic
lights switching and traffic density calculation using
video processing," 2014 Recent Advances in
Engineering and Computational Sciences (RAECS),
Chandigarh, 2014, pp. 1-6, doi:
10.1109/RAECS.2014.6799542.
[12] Ms. Saili Shinde, Prof. Sheetal Jagtap, Vishwakarma
Institute Of Technology, Intelligent traffic
management system:a Review, IJIRST 2016
[13] Arkatkar, Shriniwas & Mitra, Sudeshna & Mathew,
Tom. “India” in Global Practices on Road Traffic
Signal Control, ch.12, pp.217-242
[14] J. Hui, ‘Real-time Object Detection with YOLO,
YOLOv2 and now YOLOv3’, 2018. [Online].
Available: https://medium.com/ @jonathan_hui/
real-time-object-detection-with-yolo-yolov2-
28b1b93e2088
[15] J. Redmon, ‘Darknet: Open Source Neural
Networks in C’, 2016. [Online]. Available:
https://pjreddie.com/darknet/
[16] Tzutalin, ‘LabelImg Annotation Tool’, 2015.
[Online]. Available:
https://github.com/tzutalin/labelImg
[17] Li, Z., Wang, B., and Zhang, J. “Comparative analysis
of drivers' start‐ up time of the first two vehicles at
signalized intersections”, 2016 J. Adv. Transp., 50:
228– 239. doi: 10.1002/atr.1318
[18] ‘Pygame Library’, 2019. [Online].
Available:https://www.pygame.org/wiki/about
[19] Open Data Science, ‘Overview of the YOLO Object
Detection Algorithm’, 2018. [Online]. Available:
https://medium.com/@ODSC/ overview-of-the-
yolo-object-detection-algorithm-7b52a745d3e0

Control of Traffic Signals by AI based Image Processing

Recommended

More Related Content

Similar to Control of Traffic Signals by AI based Image Processing (20)

More from IRJET Journal (20)

Recently uploaded (20)

Control of Traffic Signals by AI based Image Processing