Kneron ModelZoo
The Kneron ModelZoo provides verified model architecture and training code for user to easily retrain model and put it on Kneron hardware platform. You can convert the retrained model to NEF model by our toolchain provided in section Kneron Toolchain Docker
In addition, the Kneron PLUS provides Generic Image Inference and Generic Data InferenceAPIs for you to quickly build the prototype application. You can learn how to leverage Generic Image Inference and Generic Data Inference APIs to do inference, pre-processing and post-processing by the following ModelZoo examples.
(*Note) There are two type of examples (Verified category/models, Legacy), recommend user can try Verified category/models first, Legacy version is planned to be removed.
Verified category/models:
| Category | Model Type & Document | PLUS example |
|---|---|---|
| kneron-mmdetection | YoloX | Python example |
| kneron-mmpose | RSN18 | Python example |
| kneron-mmsegmentation | STDC | Python example |
| kneron-mmtracking | ByteTrack | Python example |
| kneron-mmclassification | RegNet | Python example |
Legacy:
| Category | Model Type & Document | PLUS example |
|---|---|---|
| Popular Object Detection | YoloV5s | Python example |
| FCOS | Python example | |
| Popular Classification | Popular backbones | Python example |