Active Learning Algorithm for Object Detection
deep-learning
active-learning
computer-vision
object-detection
Efficient deep learning with active learning using SSD + ResNet50 on KITTI dataset
1 Overview
Research project applying active learning techniques for efficient deep learning. Minimizes labeled data requirements while maintaining effective object detection performance.
2 Tech Stack
| Category | Technology |
|---|---|
| Framework | PyTorch |
| Model | SSD (Single Shot Multibox Detector) + ResNet50 |
| Dataset | KITTI (Vehicle Object Detection) |
| Environment | Anaconda, Python 3.8 |
3 Key Features
- Random Sampling: Baseline approach for comparison
- Confidence-based Selection: Sample selection based on model prediction confidence
- Learning Loss for Active Learning (LL4AL): Advanced technique based on “Learning Loss for Active Learning” paper
- Efficient Labeling: Minimize labeled data while maximizing model performance