The scripts & jupyter notebooks in this folder provide Nexus users a guideline for those who may want to load a model for prediction and integrate them into their codebase or modify some parameters instead of using Portal to predict directly.
Each folder for the various model formats is structured accordingly:
requirements.txt: Python3 dependencies<bound_type>_<model_format>.py: Python script to load model, run inference, and overlay predictions on imagesREADME.md: Instructions for running the script
Note We recommend using a virtual environment to install the dependencies. For more information, see Python Virtual Environments.
- python 3.7 =< version
Users can currently run inference on the following model formats from Nexus in the tables below:
| Model Format | Model Architecture(s) |
|---|---|
| Tensorflow | Faster R-CNN, ResNet, EfficientDet, MobileNet, YOLOv4 |
| TFLite | Faster R-CNN, ResNet, EfficientDet, MobileNet, YOLOv4 |
| ONNX | Faster R-CNN, ResNet, EfficientDet, MobileNet, YOLOv4, YOLOv8 (New!) |
| PyTorch | Faster R-CNN, ResNet, EfficientDet, MobileNet, YOLOv4, YOLOv8 (New!) |
| CV2.DNN | YOLOv8 (New!) |
| Model Format | Model Architecture(s) |
|---|---|
| Tensorflow | Mask R-CNN |
| TFLite | Mask R-CNN |
| ONNX | Mask R-CNN, YOLOv8 (New!) |
| PyTorch | YOLOv8 (New!) |
| Model Format | Model Architecture(s) |
|---|---|
| Tensorflow | DeepLabV3, U-Net, FCN |
| TFLite | DeepLabV3, U-Net, FCN |
| ONNX | DeepLabV3, U-Net, FCN |
| PyTorch | DeepLabV3, U-Net, FCN |
Our model format support is constantly evolving. Models trained before [TBD] are not directly supported by the scripts in this folder. However, we still maintain support for these models in the legacy folder. Please refer to the legacy README for more information.