There have been many different object detection models with the YOLO prefix released in the recent years, though most of them carry a GPL or AGPL license which restricts their usage. For this reason, we selected YOLOX as the first object detection architecture since both the original code and pre-trained weights are released under the Apache 2.0 open source license.
You can find the Burn implementation for the YOLOX variants in src/model/yolox.rs.
The model is no_std compatible.
Add this to your Cargo.toml:
[dependencies]
yolox-burn = { git = "https://github.com/tracel-ai/models", package = "yolox-burn", default-features = false }If you want to get the COCO pre-trained weights, enable the pretrained feature flag.
[dependencies]
yolox-burn = { git = "https://github.com/tracel-ai/models", package = "yolox-burn", features = ["pretrained"] }Important: this feature requires std.
The inference example initializes a YOLOX-Tiny from the COCO
pre-trained weights
with the NdArray backend and performs inference on the provided input image.
You can run the example with the following command:
cargo run --release --features pretrained --example inference samples/dog_bike_man.jpg