Releases: dronefreak/dji-tello-object-detection-segmentation
Releases · dronefreak/dji-tello-object-detection-segmentation
Tello Vision Suite v2.0
Tello Vision v2.0 🚁🤖
Modern, modular instance segmentation and object detection for DJI Tello drones. Complete rewrite with SOTA models, clean architecture, and actual performance.
What Changed (Everything)
Old Stack (Legacy)
- ❌ TensorFlow 1.9 (EOL 2021)
- ❌ Python 3.6 (EOL 2021)
- ❌ Matterport Mask R-CNN (unmaintained)
- ❌ TelloPy (deprecated)
- ❌ ~4.6 FPS on 1050Ti
- ❌ Monolithic code
- ❌ Hardcoded everything
New Stack (Modern)
- ✅ PyTorch 2.0+ / TensorFlow 2.x
- ✅ Python 3.10+
- ✅ YOLOv8 (default) or Detectron2
- ✅ djitellopy (actively maintained)
- ✅ 15-30+ FPS depending on model
- ✅ Modular, extensible architecture
- ✅ Config-driven design
Features
- 🎯 Multiple Detection Backends: YOLOv8 (fast) or Detectron2 (accurate)
- 🚁 Modern Drone Control: Smooth RC controls, async processing
- 🎨 Rich Visualization: Masks, bboxes, labels with transparency
- 📹 Recording & Photos: Video recording and frame capture
- ⚙️ Config-Driven: YAML configuration for everything
- 🔌 Pluggable Architecture: Easy to add custom models
- 📊 Real-time Stats: FPS, battery, detection count