README.md

YOLOv7 supported for tfjs, tflite, saved model, openvino, onnx, rknn.
Support for WINDOWS and LINUX

Train

python train.py --batch 4 --epochs 10 --data 'data/mydataset.yaml' --cache-images disk --cfg 'cfg/yolov7-tiny.yaml' --device dml:0 --img-size 640 --weight yolov7.pt

dml:0 to set dml device (windows), cuda:0 to set cuda device, cache-images for faster training disk/ram/no

Export

python export.py --weights "yolov7.pt" --include tfjs/onnx/saved_model/openvino/tflite

for more information

python export.py --help

Usage

tfjs model can be used with makesense.ai

🔥Actions -> Run AI locally -> YOLOv5 object detection using rectangles -> and import all files from *_web_model folder

also you can follow this step to use (from https://github.com/Hyuto/yolov5-tfjs)

.onnx can be used with python.exe detect.py --weights "yolov7-w6.onnx" --nosave --source 0 --view-img 1 (use 0 to freeze screen)

References

• Official YOLOv7
• Official YOLOv6
• Official YOLOv5
• Official YOLOR
• Object Detection using YOLOv5 and Tensorflow.js
• YOLOv7 with TensorRuntime
• Make Sense