kurajulii/slamAI-istanbulCanyon
🌆 Enhance drone navigation with robust Visual Odometry and SLAM for urban canyons, focusing on realistic İstanbul environments and machine learning solutions.
What's novel
🌆 Enhance drone navigation with robust Visual Odometry and SLAM for urban canyons, focusing on realistic İstanbul environments and machine learning solutions.
Code Analysis
7 files read · 3 roundsA thin orchestration script that collects drone sensor data from AirSim, converts it to TUM/EuRoC/KITTI formats, optionally runs pre-trained ONNX models as image preprocessing, then shells out to standard ORB-SLAM3/DSO/SVO binaries and computes basic ATE/RPE metrics.
Strengths
The pipeline orchestration in main.py is reasonably structured with clear mode-based CLI arguments, and the data format conversion to TUM/EuRoC/KITTI is functional and covers the major SLAM benchmark formats.
Weaknesses
Massive code duplication between slam_integration.py and baseline_slam.py (nearly identical SLAMEvaluator classes), the 'ML enhancement' is just running ONNX models as a preprocessing step with no actual integration into SLAM algorithms, the 'modify vocabulary/config' methods are no-ops that just copy files, and the README contains fabricated download links and nonsensical installation instructions.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
8
Contributors
2
Files
15
Active weeks
5
Repository
Language
Python
Stars
1
Forks
1
License
MIT