How to Train YOLOv7 on a Custom Dataset

How to Train YOLOv7 on a Custom Dataset

How to Train YOLOv7 on a Custom Dataset

“Hot on the heels of MT-YOLOv6, a new YOLO dropped this week (and this one is a doozy).

YOLOv7 was created by WongKinYiu and AlexeyAB, the creators of YOLOv4 Darknet (and the official canonical maintainers of the YOLO lineage according to pjreddie, the original inventor and maintainer of the YOLO architecture). You can read the YOLOv7 paper here.

The model itself is impressive. Built with PyTorch, it boasts state-of-the-art performance on MS COCO for real-time object detection models (defined as running 5 FPS or faster on a V100 GPU). The various sizes of the model run at between 36 and 161 frames per second (with a batch size of one), which is extremely impressive given the high accuracy…”

Source: blog.roboflow.com/yolov7-custom-dataset-training-tutorial/

Notebook: https://colab.research.google.com/drive/1X9A8odmK4k6l26NDviiT6dd6TgR-piOa