IEEE Technology and Society Magazine - September 2023 - 91

and facilitates better analysis to combat labor
exploitation [11]. The developed MFVAS can work
in Android and IOs. The captain utilizes the MFVAS
to document the fishermen's daily or weekly working
hours. This information is then transferred to the
respective fisherman's page. To ensure accuracy, the
fisherman must log in with their credentials and verify
the working hour details using face verification
technology. The approved working hour details are
stored in an SQLite database installed on the edge
server. The SQLite database data can be transferred
to the cloud server upon landing. These recorded
working hour details are critical in analyzing and
preventing fisherman exploitation. The sentiments
regarding monitoring can vary among the captains
and the crew. Some captains may view monitoring
as a necessary aspect of ensuring the safety and efficiency
of their operations, leading to better crew
management and overall performance. They might
welcome the transparency that monitoring brings, as
it can also protect them from potential false accusations
or disputes. On the other hand, some captains
might feel uncomfortable or apprehensive about
being under constant surveillance. They may worry
about scrutinizing their actions and decisions, fearing
that it could undermine their authority or create
additional pressure in an already demanding role.
The crew's feelings toward monitoring can be complex,
particularly for migrant workers. While some
crew members may appreciate the monitoring system
as a safeguard against exploitation, harassment,
or unfair treatment, others may perceive it as intrusive
and potentially threatening their privacy and
autonomy. In any case, it is crucial to implement a
monitoring system with sensitivity and respect for
the rights and well-being of all individuals involved.
Fisherman action recognition method
On fishing vessels, the crew is typically organized
in a hierarchical structure, with the captain in charge
of the vessel and responsible for making decisions
related to fishing operations, crew management, and
safety. As with any workplace, conflicts and disagreements
can arise between crew members and the
captain due to differences in opinions, long working
hours, fatigue, or stressful situations at sea. In some
cases, disagreements may escalate to arguments,
where individuals express their differing viewpoints
or frustrations verbally. These arguments can be
intense but typically do involve physical violence.
September 2023
The crew must find ways to constructively resolve
conflicts and maintain a co-operative and safe working
environment. However, it is worth noting that
physical fighting between fishermen and the captain
is rare and not considered a normal part of the working
relationship on fishing vessels. Physical altercations
can pose serious safety risks and are generally
discouraged and unacceptable in professional settings.
Within DWF vessels, there may be instances
where conflicts arise between the fishermen and the
captain, leading to fights. It is crucial to accurately
recognize these fight actions and capture video evidence
through CCTV footage. This video evidence
is valuable in taking legal measures against fighting
among the fishermen onboard the vessels.
Convolutional neural networks (CNNs) and long
short-term memory (LSTM) networks are two types
of deep-learning models that can be used for fisherman
action recognition. A CNN is a type of neural
network that is particularly effective at learning
features from images. It uses convolution, where a
small kernel or filter is applied to the input image,
moving across it in a sliding window. The kernel performs
element-wise multiplications with the values
in the image and then sums them up, producing a
new feature map. This process is repeated multiple
times, with the resulting feature maps fed through
additional network layers. The weights of the kernels
are learned through training, allowing the network
to learn complex features from the input images.
An LSTM network, on the other hand, is a type of
the recurrent neural network (RNN) that can model
long-term dependencies in sequential data. It does
this by using a particular type of neuron called a
memory cell, which can retain information over long
periods and control when to pass it on to other neurons
in the network. LSTMs are particularly useful for
tasks that involve predicting the next element in a
sequence, such as in natural language processing or
time series forecasting.
To use a CNN and LSTM for fisherman action
recognition, preprocess the video data to extract
individual frames or clips. Preprocessed frames
or clips could then be passed through a CNN to
remove image features. The resulting features could
be fed into an LSTM network, which could then use
these features to predict the actions performed by
the fisherman in the video. Combining a CNN and
LSTM can be effective for fisherman action recognition
because the CNN can learn features from the
91
IEEE Technology and Society Magazine - September 2023

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - September 2023
