Pick-Up-That-Trash (PUTT) System

Introduction:
Waste management poses a persistent challenge, evident in the trash strewn
about in communities and schools, where directing individuals to pick up
trash can prove cumbersome.

In recent years, significant technological advancements, such as NASA's

James Webb Space Telescope and the Tesla Model Y Car with its array of
impressive features, have garnered attention. While these innovations
primarily prioritize convenience, advancement, or profitability, the emphasis
on developing technologies to address cleanliness in our surroundings is
nearly nonexistent. Presently, the available environmental technologies are
limited to sensor-activated trash bins and the use of biodegradable
materials, among other things. However, these solutions are insufficient to
meet the growing demands of waste management.

Nagi et al. (2023) proposed a system that integrates real-time object

detection using the You Only Look Once (YOLOv3) algorithm with audio
feedback to assist visually impaired individuals identify objects in their
vicinity. Building upon this concept, there exists the potential to modify the
project and create a system that automates trash collection by directing
individuals to pick up litter. This approach offers a cost-effective alternative
to purchasing complex machines. Moreover, by consistently prompting
people to pick up trash, it fosters discipline and, ideally, leads to a time
where individuals voluntarily maintain cleanliness, reducing reliance on such
systems over time.

Objectives:
The goal of this research is to develop an automated surveillance system
that actively detects trash. The following objectives are set to be used as
guides in data collection and analysis.
A. To maintain the cleanliness of the classroom.
B. To promote awareness to environmental cleanliness
C. To encourage participation in keeping the surroundings clean
This research also aims to answer the following questions:
1. Does this research help maintain the cleanliness of the classroom?
2. Can this help promote awareness to environmental cleanliness?
3. Can this encourage the students' participants in keeping the surroundings
clean?
Significance of the study:
This research can help the following:
Environment. This research will help the environment become much more
cleaner.
Future Researchers. This research will help future researchers that have a
similar topic.

Review of Related Literature:

According to Li et al. (2020), object detection is known as the core of
computer vision and it has attracted much research attention in recent years
especially because of its close relationship with video analysis and image
understanding.
Furthermore, Kaur et al. (2022) states that Object detection is a combination
of two concepts. Object classification, which defines the class of one or more
objects that exist in the image and assigns the labels of the objects and
object localization, which is a process that locates the position of one or
more objects in the image or video with the help of a bounding box. A
complete object recognition process takes an image as input, identifies the
objects, assigns labels to the object of the associated class, and gives the
class probability of the recognized object. Every object has unique features
that help to identify the class.
Additionally, Amit et al. (2021) says that each detection is reported with
some form of pose information. This could be as simple as the location of the
object, a location and scale, a bounding box, or a segmentation mask. In
other situations the pose information is more detailed and contains the
parameters of a linear or non-linear transformation.
For example, a face detector can compute the locations of the eyes, nose
and mouth, along with the bounding box of the face.
Ren et al. (2015) says that state-of-the-art object detection networks depend
on region proposal algorithms to hypothesize object locations. Advances like
SPPnet and Fast R-CNN have reduced the running time of these detection
networks, exposing region proposal computation as a bottleneck.
Acording to Sharma et al. (2017), object detection uses techniques that are
used to detect an object, localise an object, categorise an object, extract
features, appearance information, and many more, in images and videos.
Fang et al. (2019), claim that deep neural networks (DNNs) have shown
prominent performance in the field of object detection. However, DNNs
usually run on powerful devices with high computational ability and sufficient
memory, which limit their deployment for constrained environments.
Vales et al. (2022) found that the ESP32-S2 employs RTT distance estimates
different than the ones observed in the frames during FTM communications.
Although this is not detailed in the ESP32-S2 documentation, they have
experimentally verified that a linear correction is applied to the raw RTT
values depending only on the values themselves. They found that this
approach improves the accuracy in some cases but produces large errors in
other cases and overall, the FTM implementation in the ESP32-S2 is still far
from being usable in complex indoor environments. While its performance in
outdoor environments exhibit good results, with almost 90% of the samples
of absolute error for a bandwidth value of 40 MHz, in indoor environments,
only 40% of the samples fall below such a threshold.
Wahyu et al. (2022) found that the ESP32-CAM with the camera type OV2640
can be used in streaming cameras and Face detection to control the actuator.
In esp32-cam, not all types of cameras can be used, depending on the
appropriate resolution and compatibility with the camera used for image
recognition. UXVGA and SXVGA are the types of resolution with the highest
average speed (ms) >80 ms out of the nine types of resolution that the
OV2640 camera has on the ESP32-Cam. In comparison, the compatible types
for Face recognition and Image recognition are CIF and QVGA, which have a
speed of 40 ms and 19 ms.
Jiang et al. (2022) states that the core of the YOLO target detection algorithm
lies in the model's small size and fast calculation speed. They found that it
can directly output the position and category of the bounding box through
the neural network. The speed of YOLO is fast because it only needs to put
the picture into the network to get the final detection result, so YOLO can
also realize the time detection of video. YOLO also directly uses the global
image for detection, which can encode the global information and reduce the
error of detecting the background as the object. They concluded that YOLO
has a strong generalization ability because YOLO can learn highly
generalized features to be transferred to other fields. It converts the problem
of target detection into a regression problem, but detection accuracy still
needs to be improved.
Zhao et al. (2019) says that due to its powerful learning ability and
advantages in dealing with occlusion, scale transformation and background
switches, deep learning based object detection has been a research hotspot
in recent years. Their paper provided a detailed review on deep learning
based object detection frameworks which handle different sub-problems,
such as occlusion, clutter and low resolution, with different degrees of
modifications on R-CNN.
A study was conducted by Cahyono et al. (2022) where they made a
surveillance system. Their design of the security system uses the Esp32 Cam
microcontroller which is connected to the Am312 PIR sensor and the MC-38
magnet sensor. Data transmission is transmitted via the internet by utilizing
the wifi module on the Esp32 Cam microcontroller. If the smartphone is
connected to the internet network, the system can send data to the Telegram
application with only a 0.0597 second delay.

Methods and Materials:

Process Flowchart:
- Gathering the necessary materials
- Installing and developing software
- Implementing software
- Testing and fine-tuning

Procedures:
1. Gathering the necessary materials
Materials such as the ESP-32 microcontroller, FTDI module, speakers, wires,
and a 5v Lithium-ion battery are gathered.
2. Installing and training object detection software
The Tinier YOLO object detection algorithm is installed and trained to detect
trash
3. Implementing software
The now trained object detection algorithm is uploaded to the ESP-32 using
the FTDI module
4. Testing and iterating
The ESP-32, now equipped with the object detection software, is tested and
fine-tuned to perform better.