🦋 Butterfly Classification App (Species of Gujarat)

👩‍🏫 Mentorship

• Dr. Swapnil Lokhande

---

Team Members

• Smit Shah (202251122)
• Heet Shah (202251121)
• Parv Thummar (202251143)
• Tanuj Saini (202251141)

---

🔗 Links & Resources

Deployed App Links

• Mobile App: Download the Butterfly Classification App
• Access the Model directly here ->: Web App

Reports & Posters

• Final Report: View the Report
• Poster: Download the Poster

Dataset

• Dataset Link: Access the Dataset
• Dataset Curated From: iFoundButterflies

Model Checkpoints

• VGG-16 Model Checkpoints: 50 EPOCHS

SAM Models

• Github Reporitory: Repo

• SAM Model used for Project:

  • vit_l: Download vit_l (1.5GB)

• Other Available SAM Models:

  • vit_b: Download base model(0.5GB)
  • vit_h: Download large model (2.5GB)

---

📄 SAM Outputs

• SAM Model Outputs: for our dataset

📖 Introduction

Our project focuses on developing a Flutter-based mobile application designed to identify butterfly species native to Gujarat. Combining state-of-the-art AI technologies with an intuitive mobile experience, the app provides:

• Real-time butterfly identification through image segmentation and classification.
• A comprehensive catalog of 109 butterfly species with detailed information on each.
• Tools tailored for researchers, enthusiasts, and conservationists to promote biodiversity awareness and conservation efforts.

🏎 WORKFLOW

Key Highlights

• Utilizes SAM (Segment Anything Model) for precise image segmentation.
• Achieves 85% classification accuracy using a fine-tuned VGG-16 model.
• Serves as an educational and practical tool for biodiversity studies.

---

🌟 Features

• Real-time Butterfly Identification

  • Uses SAM for precise segmentation and VGG-16 for accurate classification.
  • Supports multiple image formats like JPEG, PNG, and JPG.

• Species Catalog

  • A catalog of 109 butterfly species native to Gujarat, showcasing detailed information for each.

• User History and Profile

  • Allows login/signup functionality to maintain a personalized user experience.
  • Tracks identification history for future reference.

• Direct Online Search

  • Clickable links for each species enable instant searches for additional information on Google.

---

🧑‍💻 Methodology

1. Data Collection and Preprocessing

• Collected images of butterfly species from Gujarat from reliable online sources.
• Applied preprocessing techniques to optimize the dataset for training.

2. Model Selection

• Tested multiple architectures, including ResNet and DenseNet.
• Selected VGG-16 for its superior performance, achieving 85% accuracy.

3. Image Segmentation

• Implemented SAM to isolate butterflies from their surroundings.
• Enhanced classification reliability by focusing on segmented objects.

4. Flutter App Development

• Designed and developed a user-friendly app with key functionalities:
  • Login/Signup
  • Species catalog
  • History tracking

---

🔧 Tech Stack

Component	Technology Used
Frontend	Flutter
Backend	Mongo DB
Image Segmentation	SAM (Segment Anything Model)
Classification Model	VGG-16 (Fine-tuned)

---

📈 Future Scopes

• Expand the catalog to include national butterfly species beyond Gujarat.
• Develop a community-verified and research-backed repository for consistent updates.
• Incorporate local sightings and academic contributions to enhance the app's richness.

---

📸 Screenshots and Visuals

Masking BY SAM Model

📅 Future Enhancements

• Expand to National and International Species:
  • Increase the app's reach by adding species from different regions and countries.
• Community Contributions:
  • Create a platform for users to contribute new species sightings and academic content.

---

This Butterfly Classification App is an excellent tool for promoting biodiversity awareness, providing researchers and enthusiasts with a valuable resource for identifying and learning about butterfly species in Gujarat. 🌿

---