Skip to content

API Changes

Navigation

TOC is limited to main sections for better readability. Use Ctrl+F to search for specific changes.

Change Date Impact Breaking Status
YOLOv5 to YOLOv8 Upgrade Aug 2025 Low ❌ No ✅ Complete
Removed CARLA Version Argument Aug 2025 Medium ✅ Yes ✅ Complete
Version Import Changes Aug 2025 Low ✅ Yes ✅ Complete

Machine Learning Models

What Changed: Upgraded from YOLOv5 to YOLOv8 for object detection with automatic fallback

Component Status Replacement/Improvement
YOLOv5 object detection ✅ Supported YOLOv8 with YOLOv5 fallback
torch.hub loading ✅ Supported ultralytics package (preferred)
Detection output format ✅ Compatible Automatic format conversion 
# Your existing code works unchanged
from opencda.customize.ml_libs.ml_manager import MLManager

ml_manager = MLManager()  # Automatically uses YOLOv8 if available
detection = ml_manager.object_detector(images)
  • Better Performance: YOLOv8 is ~20% faster than YOLOv5
  • Higher Accuracy: Improved detection accuracy on COCO dataset
  • Modern API: Uses ultralytics package (already in dependencies)
  • Thread Safety: Fixed GIL threading issues with cv2.waitKey
  • Automatic Fallback: Falls back to YOLOv5 if YOLOv8 unavailable
# Automatic model selection
try:
    from ultralytics import YOLO
    self.object_detector = YOLO('yolov8m.pt')  # YOLOv8 medium
    self.use_v8 = True
except ImportError:
    self.object_detector = torch.hub.load('ultralytics/yolov5', 'yolov5m')
    self.use_v8 = False

# Output format compatibility maintained
# YOLOv8 results automatically converted to YOLOv5 format for camera-lidar fusion

Impact: None - fully backward compatible
Benefits: Better performance, improved accuracy, modern codebase, thread safety

Command Line Interface

What Changed: Removed -v/--carla_version argument, CARLA version now fixed to 0.9.15

API Status Replacement
opencda.version module ❌ Removed from opencda import __version__
Command line -v argument ❌ Removed Fixed to CARLA 0.9.15 
# Before (will fail)
python opencda.py -t single_2lanefree_carla -v 0.9.12
python opencda.py -t platoon_joining_2lanefree_carla -v 0.9.11

# After (required)
python opencda.py -t single_2lanefree_carla
python opencda.py -t platoon_joining_2lanefree_carla
  • Remove -v or --carla_version from all scripts
  • Update CI/CD pipelines
  • Update documentation examples

Impact: All command line scripts need updating Benefits: Simplified interface, consistency across scenarios, reduced compatibility issues

Function Signatures

What Changed: Version import moved from separate module to main package

# Before (will fail)
from opencda.version import version

# After (required)
from opencda import __version__, CARLA_VERSION

from opencda import __version__, CARLA_VERSION

print(f"OpenCDA version: {__version__}")      # "0.1.3"
print(f"CARLA version: {CARLA_VERSION}")      # "0.9.15"

Impact: All code importing version information needs updating
Benefits: Cleaner package structure, centralized version management

Compatibility & Stability

Stability Commitment

  • Core OpenCDA APIs: Remain stable with backward compatibility
  • MARL APIs: Experimental, may change in future versions
  • Breaking Changes: Always documented with migration guides
Component Stability Versioning
Core OpenCDA ✅ Stable Semantic versioning
MARL Extension ⚠️ Experimental May change
CLI Interface ✅ Stable Documented changes

← Back to Changelog