IR Carbide Canon

Human wild conflicts happen in Kerala very often. Hundreds of lives and damage to agriculture happen every year due to human wild conflicts. The main objective of this project is to reduce the human wild conflict with the help of AI and actuator mechanism. The setup uses Xilinx kria KV260 vision AI starter kit. The edge AI capabilities of KV260 are used to identify the type of wild animal encroaching the area. A suitable labeled data set is identified to develop the machine learning model based on CNN. Once the wild animal is identified, a suitable scare mechanism will be actuated. The different types of actuators include high-intensity flashlights, bee sounds, carbide-based actuators, ultrasonic sounds etc. We faced several issues while developing the system. One of the major issues was the integration of the ML model into the KRIA starter kit.

Procedure:

We used PetaLinux in this project - Kria KV260 AI starter Image

The image is flashed into the sd card using Balena etcher

Set up the Kria KV260 kit as per the instructions obtained from the starter guide (Insert the SD card, power adapter, and camera)

Update all the packages using the below code

sudo xmutil getpkgs

Install smart camera

sudo dnf install packagegroup-kv260-smartcam.noarch

Load smart camera

sudo xmutil listapps
sudo xmutil unloadapp
sudo xmutil loadapp kv260-smartcam
sudo smartcam --mipi -W 1080 -H 720 -r 30 --target dp

The smart camera application provides face detection.

Our aim is to integrate the wild animal detection model to Kria KV260 and based on the detection a suitable actuator will be triggered.

We faced issues in the following and is trying to rectify the same

- After repeated trials, the starter kit fails to load the hardware accelerator

- model creation using vivaldo failed and was not able to integrate the model with the kit