repository19/Real-Time-Object-Detection-with-Depth
🔍 Detect objects in real-time and extract their 3D coordinates using YOLOv5 and Intel RealSense D435 for robotic arm tasks.
What's novel
🔍 Detect objects in real-time and extract their 3D coordinates using YOLOv5 and Intel RealSense D435 for robotic arm tasks.
Code Analysis
11 files read · 4 roundsA YOLOv3-based object detection system that computes stereo depth via SGBM matching on a hardcoded stereo camera pair, but does not actually integrate the depth information with the detection results.
Strengths
The YOLOv3 detection pipeline is complete and well-structured with proper training (freeze/unfreeze, GIoU loss, data augmentation), inference, and evaluation (mAP computation). The stereo depth pipeline (rectification, SGBM, 3D reprojection) is correctly implemented for the specific camera rig.
Weaknesses
The README falsely claims YOLOv5 and Intel RealSense D435 when the code is YOLOv3 with a generic stereo camera pair. The depth computation is computed but never actually linked to detected objects — the core promise of 'detecting objects in 3D space' is not fulfilled. Camera parameters are hardcoded, there are zero tests, and the SGBM pipeline is duplicated across files.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
20
Contributors
2
Files
39
Active weeks
4
Repository
Language
Python
Stars
3
Forks
0
License
—