Open tohruk opened 3 years ago
Hi! I believe YOLOv3 and YOLOv3-tiny are the same in terms of input dimension (both work with 608x608 or 416x416). The difference is in model architecture: YOLOv3-tiny has lesser convolution layers than YOLOv3, hence speeds up detection.
README.md says that this benchmark script supports "YOLO-V3 : 608x608", but only YOLO-V3 Tiny is available. Is there any option to perform a benchmark for YOLO-V3(608x608 or 416x416)?