FishSense iOS Application

This project simplifies fish measurement and monitoring during catch-and-release operations. We’re developing an iOS app that uses the device’s camera and LiDAR to detect fish, identify head and tail positions, and calculate length automatically. With machine learning, the app will also identify species and streamline data collection for marine or citizen scientists.

Key Features

• Accurate Fish Length Detection – Uses the iOS device’s camera and LiDAR.
• Fish Species Detection – Applies machine learning to recognize species.
• Session Management & Cloud Upload – Organizes and uploads data efficiently.

Repository Organization

We use two repositories. Follow the README in each for replication:

• iOS Swift App

  Swift code for real-time fish detection using the iPhone's camera and LiDAR.

  See the README for setup and usage.

• fishsense.rs

  Rust library for fish length estimation from sensor data.

  See the README for build and integration steps.

System Design

Application

Sample screenshot of the app in action measuring fish length:

Analysis of Fish Length Estimation Error

Our algorithm has an error rate of less than 10% on iPhone pro and kess than 16% on iPad pro, demonstrating high accuracy.

For more details, refer to the final report below.

Fish Species Classification Model

We use a classification model based on Fishial, exported from TorchScript to ONNX for efficient iOS deployment via ONNX Runtime.

Model available on Hugging Face:
ONNX Fish Classifier

This enables real-time, on-device fish species recognition.

Team Members

• Anushri Eswaran – aeswaran@ucsd.edu
• Ziyuan Lin – zil102@ucsd.edu
• Tanishq Rathore – trathore@ucsd.edu
• Chris Crutchfield – ccrutchf@ucsd.edu

Documentation & Resources

• Milestone Report
• Project Specification
• Demo Videos

Contact & Links

• iOS Repository: FishSenseSwift
• Rust Repository: fishsense.rs
• For questions, open an issue or email a team member.