Idea
A semi-supervised parking slot detection platform using large-scale real-world data to enhance autonomous parking system accuracy.
Research Paper
Core Innovation
This paper presents CRPS-D, the largest and most diverse parking slot detection dataset capturing real-world noise and varied conditions. It also proposes SS-PSD, the first semi-supervised parking slot detection model that improves performance by exploiting unlabeled data through a teacher-student framework with confidence-guided mask consistency and adaptive feature perturbation.
Market Size (TAM)
$10–20B TAM for autonomous vehicle perception systems; $2–10B SAM from automotive OEMs and smart parking solution providers. Driven by increasing adoption of autonomous parking and smart city infrastructure.
Potential Customers & Pain Points
- Automotive manufacturers developing autonomous parking systems
- Parking management companies needing accurate slot detection
- Smart city planners integrating parking solutions
- AI researchers requiring large-scale annotated datasets
- Autonomous vehicle software developers facing noisy real-world data challenges
Business Model
Open-source dataset and baseline model with licensing for commercial use; subscription or licensing fees for enhanced models and enterprise support; consulting for integration with automotive and smart city platforms.
Competitive Landscape
- Tesla Autopark
- Waymo Parking Solutions
- Mobileye Parking Detection
Implementation Challenges
- High cost and complexity of large-scale real-world data annotation
- Integration challenges with diverse vehicle sensor systems
- Robustness under extreme environmental conditions
Validation Strategy
- Benchmark SS-PSD against state-of-the-art models on CRPS-D and public datasets
- Pilot integration with automotive OEMs for real-world autonomous parking tests
- Collect feedback and iterate model improvements based on deployment data
Research Paper Overview
Advancing Real-World Parking Slot Detection with Large-Scale Dataset and Semi-Supervised Baseline
Summary
This study introduces CRPS-D, a large-scale parking slot detection dataset with diverse real-world conditions and a semi-supervised baseline model SS-PSD that leverages unlabeled data to improve detection accuracy. The approach uses a teacher-student framework with confidence-guided mask consistency and adaptive feature perturbation, outperforming existing methods on multiple datasets. The dataset and code are publicly available.