Professional Documents
Culture Documents
The Making of Object Recognition Eyeglasses For The Visually Impaired Using Image AI
The Making of Object Recognition Eyeglasses For The Visually Impaired Using Image AI
Abstract:- People with visual impairment may face In order to alleviate the challenges faced by visually
struggles in their daily activities, as these may affect impaired individuals, recent advancements that are
them socially, physically, and psychologically. This study embedded in systems and artificial intelligence need
aims to address this problem by utilizing quantitative development. Thus, devices like smart glasses were
experimental research to come up with Object manufactured by Google Glass, wherein the detection
Recognition Eyeglasses out of ImageAI. This device aims process is transformed by signaling the visually impaired
to assist the visually impaired person by recognizing the through an audio alert or vibration (Mukhiddinov& Cho,
object in front of them and giving an audio output of the 2021). These smart glasses can distinguish objects and
name of the object. Throughout the testing process, the signals when confronted with various difficulties; however,
Object Recognition Eyeglasses showed accuracy in the smart glasses can only detect at proximity determined by
recognizing different objects and their different signal strength. Additionally, a device that also functions as
varieties. It also showed its capability to recognize an assistive device is a guide cane, which works as a
objects from far distances, with a maximum distance of mobility aid to identify objects and employs tapping sounds
45 meters, and its efficiency in delivering a timely for echolocation, similar to the companion cane (Hersh,
recognition and audio output with an average time 2015). In contrast, these guide canes are inaccurate and
interval of 1.61 and 0.63 seconds respectively. Based on unreliable when it comes to navigating and approaching
these results, the Object Recognition Eyeglasses stands dangerous areas, and present certain issues for visually
as an accurate, efficient, and capable assistive device that impaired people. Similarly, another study presented Object
can help visually impaired people in their day-to-day Recognition Glasses with a fusion of Raspberry Pi and Pi
lives. However, this device still needs improvement in camera for intricate facial recognition procedures, including
terms of convenience by using a phone instead and its detection utilizing an ultrasonic sensor, wherein this
modifying it to not require any internet connection. technology significantly enhanced individuals' quality of life
(Dematti et al, 2023). The studies covered various
Keywords:- Object Recognition Eyeglasses, Image AI, prototypes that provided effective and secure assistance in
Visually Impaired, Assistive Device locating unfamiliar objects and improving their mobility.
How long is the time interval of the Object Recognition Below is the Testing Procedure to Measure the Time
Eyeglasses between recognizing the object and playing Interval of the Object Recognition Eyeglasses between
the sound of the name of the object? Seeing the Object and Recognizing the Object
IV. RESULTS
The section shows the results and interpretation of data that were collected from assembling and testing the device.
The accuracy of the Object Recognition Eyeglasses B. The Maximum Distance of Recognition of the Object
was tested using different samples of different objects and Recognition Eyeglasses:
testing whether it can identify each sample.
Table 2: The Maximum Distance of Recognition
The data on Table 1 displayed the percentage of Distance Recognition
accuracy of the Object Recognition Eyeglasses in 10 meters Yes
identifying the various objects in the vicinity. Each trial has 20 meters Yes
five distinct samples of varying shapes and sorts to 30 meters Yes
determine its correctness with a “YES” response expected in 40 meters Yes
each case. A bottle in the shape of a short plastic bottle, a 45 meters Yes
glass bottle, a flask bottle, or a plastic tumbler is the first 50 meters No
example item. A potted plant with varying components
served as the second example object. Three various types of The researchers tested the maximum distance the
cups—a glass cup, a paper cup, and a mug—were included Object Recognition Eyeglasses can recognize an object by
in the third example item. Bicycles of various sizes and sorts using the same object, in this case, the researchers used a
make up the fourth example object. The last example item person, and tested whether it can recognize the person at a
used to evaluate the assistive device's accuracy was a set of continuously farther distance.
colorful keyboards. With a high success rate of ”YES”, the
Object Recognition Eyeglasses identified all five sample Table 2 displays the greatest distance that the Object
objects—bottles, cups, potted plants, bicycles, and Recognition Eyeglasses can recognize an object in meters.
keyboards—in all of their different shapes and varieties. The recognition range was tested at six different distances to
ensure accuracy. The distances in the list were accurately
Furthermore, related research employed Tensorflow or computed using a measuring tape as a reference. The AI
Python Open pre-training to effectively identify objects correctly identified the person in front of the camera at a
using a camera. The smart glasses improved performance in distance of 10 meters and said "Person." The following
identifying the item in front by 50–60% (Shakkir, 2022). distances apply as well: 20, 30, 40, and 45 meters. After
The equipment works well and can enable those who are testing the AI, the results showed that it could reliably
blind or visually impaired to walk and navigate on their distinguish the person standing in front of the camera at
own. Another comparable gadget was made to alert the user. distances ranging from 10m to 45m. At 50m, the AI was
Users of augmented reality smart glasses can have a more unable to identify the person since they were too far away
remarkable immersive experience thanks to the display or from the camera.
projection system's usage of sensors to track things in view
(Koutromanos&Kazakou, 2023). The ABGs' use of sensors In a similar investigation, accurate photographs were
lowers the challenges that people who are visually impaired taken at a distance estimated to be between 40 and 150
have when doing everyday tasks. meters using an ultrasonic sensor, which generates
ultrasonic waves and detects their reflections (AlSaid et al.,
s=sample 2019). In a similar vein, a study conducted to determine
distance for visually impaired people utilizing object
detection used an HQ image camera with a LIDAR sensor to
provide high-precision distance measurements with accurate
object recognition (Dragne, 2022). Both experiments
demonstrated that the assistive device is efficient and C. The Time Interval of the Object Recognition Eyeglasses
effective. between Seeing an Object and Recognizing it
The time interval of the Object Recognition Eyeglasses
was tested between seeing an object and recognizing it by
using a stopwatch to measure the time it takes from the
moment the Object Recognition Eyeglasses sees the object
until the program is able to recognize the object.
Table 3: Time interval of the Object Recognition Eyeglasses between seeing an object and recognizing it
Trial 1 2 3 Average
Time interval 1.83 seconds 1.58 seconds 1.41 seconds 1.61 seconds
(in seconds)
Table 3 presents the three different trials done from recognition eyewear discovered that the system could
seeing to recognizing the object by the Object Recognition identify objects and barriers and warn visually impaired
Eyeglasses by using a stopwatch. The average was people of impending dangers (Kumar et al., 2019). In a
computed by adding all the results and dividing the sum by related investigation, the performance of the glasses was
the number of trials. Testing with a pair of scissors, the first evaluated in terms of recognition time using two metrics that
trial succeeded in seeing and recognizing it with a delay of demonstrate improving techniques that lead to an
1.83 seconds. In the second trial, the camera succeeded; It improvement in accuracy and time. Elnabawy et al., 2022).
saw and recognized it with a delay of 1.58 seconds. In the
third trial, the camera succeeded; It saw and recognized it D. The Time Interval of the Object Recognition Eyeglasses
with a delay of 1.41 seconds. between Recognizing the Object and Playing the Sound
of its Name
According to the results, it took the AI, on average, The time interval of the Object Recognition Eyeglasses
1.61 seconds to correctly recognize and identify the name of was tested between recognizing an object to playing the
the object that was given. The three trials demonstrate that sound of its name by using a stopwatch to measure the time
there is a time-dependent lag in seeing and identifying the it takes from the moment the Object Recognition Eyeglasses
provided object because the AI frequently detects additional recognizes the object until the program is able to play the
objects behind the object that the camera has captured. sound of its name.
Another research that used a Raspberry Pi to create object
Table 4: Time Interval of the Object Recognition Eyeglasses between Recognizing an Object and Playing the Sound of its Name
Trial 1 2 3 Average
Time interval 0.66 seconds 0.68 seconds 0.56 seconds 0.63 seconds
(in seconds)
Table 4 depicts the three trials for determining the time The following is the summary of results for each
interval between recognizing and playing the object's sound Statement of the Problem of this study.
using a stopwatch. Three trials were undertaken to
determine the data's credibility, and the average was Accuracy of the Object Recognition Glasses in
computed by dividing the sum of all values by three. In the Identifying Objects
first trial, the audio played for an average of 0.66 seconds.
In the second trial, the audio played for an average of 0.68 Five sample objects of varying types and forms were
seconds. And lastly the third trial, the audio played in an tested, the glasses accurately identified the objects,
average of 0.56 seconds. which resulted in all of the samples being marked as
“YES” and a high percentage of success rate.
According to the data, the Object Recognition
Eyeglasses took an average of 0.63 seconds to play the Maximum Distance the Object Recognition Glasses can
audio after detecting an object. The tests revealed that the Recognize an Object in Meters
audio of the Object Recognition Eyeglasses had the smallest
latency in playing the sound. To guide the vision impaired, The glasses were tested at six distinct distances.
the ImageAIactivates audio and plays it after identifying an Beginning at 10 meters, reaching the maximum distance
object to avoid accidents. The glasses have been shown to at 45 meters, and by 50 meters, it could no longer detect
translate visuals into audio in 3 seconds. Similarly, many the particular object.
smart glasses systems have Artificial Intelligence that
incorporates audio feedback to deliver real-time solutions in
navigating, such as a similar program with the use of eSpeak
was integrated into the Object Recognition Eyeglasses,
wherein the convenience of the audio's speed helps the
visually impaired locate objects faster (Islam et al., 2023).
Time Interval of the Object Recognition Glasses between Object Recognition Eyeglasses make people aware of
Seeing an Object and Recognizing it the struggles of the visually impaired encounter. Visually
impaired people will have an enhanced and safe lifestyle
The glasses were tested in three trials to determine the while doing simple tasks, wherein their quality of life
time interval between seeing and recognizing the object will be improved and at the same time ensure their
using a stopwatch. On Trials 1, 2, and 3, the glasses wellness.
recognized the sample object with a delay of 1.83 Moreover, future researchers may use this study to assist
seconds, 1.58 seconds, and 1.41 seconds, respectively. in developing a project with similar components. Future
researchers may include more detectors and speed in
The Time Interval of the Object Recognition Glasses playing the audio. The assistive devices should expand
between Recognizing the Object and Playing the Sound the distance to locate other present objects. The research
of its Name suggests using a phone instead of a laptop in
programming the software to make it more convenient.
The voice-activated Object Recognition Glasses The research also suggests making the device eliminate
produced a fast rate of playing audio with a fastest time the Wi-Fi connection. The ability to focus on the given
of 0.56 seconds with a minimum delay of 0.63 seconds object instead of recognizing other objects at once can
in recognizing the object to play the name of the object also be further tested. Future researches are also
presented. suggested to improve the number of objects that can be
recognized due to only recognizing 50 different types of
V. CONCLUSIONS objects.