YOLOv7.jl

This package is a Julia version of the YOLOv7 model based on the original author implementation in PyTorch.

To Do

There is still some things to be done, collaboration is appreciated from anyone.

• Compare with Python YOLOv7
• Add basic training loop
• Add data loading
• Add data preprocessing and augumentation
• Add more models

And more...

Inference

YOLOv7.jl provide one constructor that loads pre-trained weights that were made available by the original author implementation in PyTorch.

Currently we only support loading the "standard" version of YOLOv7.

using YOLOv7, Flux

model = yolov7_from_torch()
testmode!(m, true)

#
#   LOADING DATA ...
#

ŷ = model(x)

You can check some inference tests made using these weights on some images: