S

Segformer B5 Finetuned Coralscapes 1024 1024

Developed by EPFL-ECEO
SegFormer model optimized for coral reef semantic segmentation tasks, fine-tuned on the Coralscapes dataset at 1024x1024 resolution
Downloads 821
Release Time : 3/21/2025

Model Overview

This model is a semantic segmentation model based on the SegFormer architecture, specifically designed for scene understanding in coral reef ecosystems, capable of accurately identifying and segmenting different semantic categories in coral reef images.

Model Features

High-Resolution Processing Capability
Supports 1024x1024 high-resolution image input, suitable for fine segmentation of coral reef scenes
Data Augmentation Optimization
Employs various data augmentation strategies during training, including random scaling, rotation, and color jittering, to enhance model robustness
Sliding Window Prediction
Provides sliding window prediction functionality, capable of processing input images of any size

Model Capabilities

Coral Reef Image Semantic Segmentation
High-Resolution Image Processing
Ecological Scene Understanding

Use Cases

Ecological Monitoring
Coral Reef Health Assessment
Evaluates the health status of coral reef ecosystems by segmenting different biological categories in coral reef images
Can identify 40 different coral and marine life categories
Underwater Ecological Survey
Automatically analyzes coral coverage and distribution in underwater photography or videos
Achieves an average IoU of 57.8% on the test set
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase