YOLOv3-Tiny — 2-scale lightweight variant (Redmon & Farhadi, 2018).
Builds the YOLOv3-Tiny detector explicitly defined in the paper's accompanying release: shallower Darknet-Tiny backbone with only two detection scales (strides 16 and 32) and 3 anchors per scale. Approximately 8.7M parameters; targets edge / real-time deployment.
Model Size
Parameters
pretrainedbool= FalseIf
True, attempt to load pretrained COCO weights. Currently
raises NotImplementedError.**overridesobject= {}Keyword overrides forwarded into
YOLOV3Config.Returns
_YOLOV3TinyDetector with the YOLOv3-Tiny configuration applied (or with
overrides merged on top of it).
Notes
See Redmon & Farhadi, "YOLOv3: An Incremental Improvement", 2018 (arXiv:1804.02767). Drops one detection scale and uses a much smaller backbone — the standard "tiny" variant ships alongside the full release for embedded inference.
Examples
>>> import lucid
>>> from lucid.models.vision.yolo._v3 import yolo_v3_tiny
>>> model = yolo_v3_tiny()
>>> x = lucid.randn(1, 3, 416, 416)
>>> out = model(x)
>>> out.logits.shape[0]
1