Sovereign 8s.is mirror of standardgalactic/Class-Incremental-Few-Shot-Underwater-Object-Detection-Framework
- Python 100%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
| fs | ||
| utils | ||
| README.md | ||
| test_fasterrcnn.py | ||
Class-Incremental-Few-Shot-Underwater-Object-Detection-Framework
To this end, this paper proposes a class-incremental few-shot underwater object detection framework(CIFS-UD) This repository provides the official implementation of CIFS-UD, a Class-Incremental Few-Shot Underwater Object Detection framework designed to address catastrophic forgetting and data scarcity in underwater object detection.
CIFS-UD
Class-Incremental Few-Shot Underwater Object Detection
📌 Overview
CIFS-UD is a Class-Incremental Few-Shot Underwater Object Detection framework designed to address:
- Catastrophic forgetting in class-incremental learning
- Data scarcity in underwater environments
- Prototype-guided incremental adaptation
The framework adopts a three-stage training paradigm:
- Base Training
- Prototype Extraction
- Incremental Few-Shot Fine-Tuning
🗂 Dataset Preparation
Supported datasets:
- Brackish
- TrashCan
- RUOD
All datasets must be converted to Pascal VOC 2007 format.
Directory structure:
⚙️ Installation
Requirements
- Python >= 3.8
- CUDA 11.8
- PyTorch 2.0.0
Install PyTorch:
pip install torch==2.0.0+cu118 torchvision==0.15.1+cu118 --index-url https://download.pytorch.org/whl/cu118
Evaluation
python test_fasterrcnn.py --config-file configs/RUOD/split1/inc/10shot_INC_CLP.yml --eval-only