HE-5 | Devpost

Hardware Prototype version 1
This photo shows the first step simulation of our Prototype like how we are integrating the sensors for our aim .
This is the updated version 2 of the simulated prototype in which we have added the Seismographic transducer for Underground activities.
An example of FLIR Thermal Dataset
Setup of Communication between the Microcontroller Interface and Receiver.
Output of the Seismographic Transducer

Inspiration -Though Border Security forces striving hard to prevent infiltration across Borders , Smart Systems can be deployed across the borders which will be a helping hand for the border security forces to prevent the suspicious activities at LC.

What it does- Our Prototype will showcase the Smartisation model of AIOS which includes sensors like IR , seismographic transducers , Microcontroller and Servo Motors along with Thermal Imaging Camera . The problem statement is worked out by dividing it into two parts -

Detection and Identification. By Using the Sensor Integration Techniques the suspicious activities of enemies at LOC will be detected and after that using the Image Detection Techniques employing Artificial Intelligence to identify the Intruder and further Action can be taken.

How we built it- We divided our Model Development in two parts one is Hardware and other one is Software. Further Hardware part is divided into two parts-

Simulation Part- We have done our circuit simulation on Tinker-CAD platform which contains all the required Equipment's and Sensors.
Hardware Circuit Design - It contains the array of three IR Sensors and Servo Motor such that based on the feed of IR through microcontrollers actuates the Servo in particular Direction. We have used the Arduino (Microcontroller) for our Purpose. Now by the movement of Motor the thermal Camera Mounted on it will also rotates and the work now transfers to Software domain after Capturing the Frames. For making the model we are using FLIR Thermal Dataset which is available free of cost on internet the annotations were in coco format so it's needed to convert the COCO annotation to Yolov4 Annotations. Tuning Tiny Yolov4 framework for training and the model will be trained based on dataset and it will be ready for Identification. After the Intruder has been Identified the Signal is transferred to control center via the underground optical fibers so the alarm gets triggered and necessary action can be taken.

Challenges we ran into- The following are the challenges we faced till now -

1) The greatest challenge we faced was to figure out how to carry the simulations for Electronics side and AI side. We were trying to use different simulating Software's and found Tinker CAD for design purpose and tiny yolov4 for AI. On the other hand, finding the data set for Object detection was quite tough and to test it. 2) We finally made a working simulation using Tinker CAD and got the output and the simulation using tiny yolo v4 was also successful. These accomplishments we have made will really help us making a good and working prototype.