研究テーマ
テー
マ:Recognition of Driving Environment Using Deep
Learning
担当:
G19601 Adhikari Ashish
Ashish Adhikari(修士2年)
概要:
In recent
years, there are a bunch of autonomous driving
technologies to help drivers on road. These
technologies are highly expensive for ordinary
people to afford. Technologies like road
condition recognition, surrounding recognition,
speed control notifications, sudden acceleration
prevention, signal detection, and notification
technologies are available in the market. Most
of those systems come with the automobile
itself. In this research, we propose an image
analysis-based smartphone application model and
the dataset to train. We used the Caffe
implementation of the MobileNet-SSD detection
network, with pre-trained weights on Visual
Object Classes Challenge 2012 (VOC2012)
with the dataset we prepared. Counting depthwise
and pointwise convolutions as separate layers,
MobileNet has 28 layers. Android terminals were
used to demonstrate the model efficiency and the
experiment was conducted in the daytime on the
public roads as well as by showing the screen.
Although, the system was not fully efficient.
Analysis and detection tasks validation was
performed manually as well to avoid mechanical
errors. While validating misrecognized and
undiscovered objects, images are manually taken
for re-train. As a result, the mean average was
0.712. The upper level was the farthest so the
detection was also the lowest as 0.370, the
middle level was 0.889 and the lower past was
0.750 in the single-lane road. Double lane road
was 0.458, 0.944, 0.857 for upper, middle, and
lower levels respectively.
テーマ:Mental
State Classification by Using Brainwave Sensors
担当:Sukh Sagar Subedi(修士2年)
概要:
In recent
researches, BCI (Brain-Computer Interface) is
being pulled as a research topic. Development of
BCI can be applied in various places from home
to medical sectors. Basically, BCI is the
interconnection of human brain and computer. BCI
is the communication pathway between the
external peripheral devices. We classify our
research in two parts. Firstly, we tried to
control the IoT device named MaBeee with the
help of Neurosky Mindwave EEG sensor. The
combination of these two different sensors is
the first approach among the researchers. In
another research, two different brainwave
sensors Neurosky Mindwave Mobile and Emotiv
EPOC+ sensors have been used. The purpose of
this research is to analyze the brainwave data
for the best accuracy result and also to compare
the accuracy result between both sensors.