Project Report -> https://drive.google.com/file/d/1uf190dMKS53QddYgG-PH6x-7NyU5RKab/view?usp=drive_link
Pranav T Selven
Atiq Urrahman
Vijay Joshi
Sahil Goyal
API link - >http://ec2-13-49-46-41.eu-north-1.compute.amazonaws.com/docs
Website Link -> https://remx-web-flask.onrender.com
Model Development Repo-> https://github.com/Vijay-J0shi/remx_yolo
Flask Website Development Repo-> https://github.com/Vijay-J0shi/Remx_Web_Flask
React Website Repo (Developed By atiq and Adam) ->https://github.com/Adam-Al-Rahman/remx-website
REMX API is an automated REM (Random Encounter Model) tool for animal abundance estimation using Machine Learning
Remx is an innovative API designed to automate the detection and measurement of wildlife in India, integrated as a component of the Animeter application used by Wildlife India. The project leverages motion-sensing cameras placed in wildlife habitats to capture thousands of images triggered by animal presence. These images are processed to identify animals and generate bounding boxes around them, enabling precise measurements based on a pixel-to-distance ratio. Traditionally, this process required manual annotation of bounding boxes across vast datasets, a labor-intensive and time-consuming task. Remx automates this process using a YOLOv8-based deep learning model, delivering accurate bounding box predictions through a RESTful API. This API is designed for seamless integration into the Animeter software, developed by a senior team, to enhance wildlife monitoring and conservation efforts.
The primary purpose of Remx is to streamline the process of wildlife monitoring by automating the detection and measurement of animals in images captured by motion-sensing cameras. By providing an API that predicts bounding boxes around animals, Remx eliminates the need for manual annotation, significantly reducing the workload for wildlife researchers and conservationists. The project aims to:
- Enhance the efficiency of data processing for wildlife monitoring.
- Provide accurate and reliable bounding box predictions for animal measurements.
- Support conservation efforts by enabling scalable analysis of wildlife populations and behaviors.
- Integrate seamlessly with the Animeter application to provide a robust tool for Wildlife India.
- Model Training: Fine-tunes a YOLOv8 model and exports it to ONNX for deployment.
- Image Processing: Accepts single images or ZIP files, preprocesses them, and predicts bounding box coordinates using YOLOv8.
- API Endpoints: Provides a RESTful endpoint (/api/predict/upload) for uploading files and returning predictions.
- Web Interface: Enables users to upload files, view predictions, download CSV results, and access historical data.
- Data Persistence: Stores prediction results in a SQLite database for historical analysis.
- CI/CD Pipelines: Automates testing, linting, and deployment for both API and web application.
REMX API offers two licensing options to accommodate diverse use cases:
- GPL-3.0 Licence: This OSI-approved open-source license is ideal for students and enthusiasts, promoting open collaboration and knowledge sharing. See the GPL-3 License file for more details.