Experiment Details & Features
These interactive dashboards display results iteratively collected for the latest Ultralytics YOLOv26 objects detection model under various configurations. The graphs support zooming, panning, saving outputs, and trace isolation (by clicking on legend items to toggle visibility).
- Architectures: regular YOLO26, plus P2 and P6 variants
- Training Image Sizes: 512, 640, and 800
- Inference Precisions: FP32 and Half-Precision
- Inference Image Sizes: 640 and 800
- Pretraining Strategies: Fully trainable parameters, and pre-trained with frozen backbone.
- Model Sizes:
n, s, and m