Note
LEIP supports many additional deep learning models. The models listed below only represent models that are supported for use in LEIP Design (Golden Recipes and Recipe Designer) and contain additional optimization and performance testing.
If you would like to use LEIP with additional models via Bring Your Own Model (BYOM), see Model Preparation and Supported Formats
Future releases will support a greater range of ONNX models, including transformers.
Below is a list of our currently supported models for use in LEIP Design, whether you’re Bringing Your Own Data (BYOD) to use with a Golden Recipe or creating your own recipe.
Classifier Models
LEIP supports most of the pytorchcv and timm models.
Detector Models
LEIP currently supports the following detector model families:
- EfficientDet
- YOLOv5
- YOLOv8
- SSD
- NanoDet
Best Performing Detector Backbones
The performance of the supported detector model families was tested extensively and used to generate the Golden Recipe Database (GRDB). More details (target architecture, inference speed, accuracy, etc.) about models that passed LEIP end-to-end testing can be found at the GRDB Explorer.
The following detector backbones were in the top 0.75% of the GRDB by performance (mAP):
| Model Family | Backbone |
|---|---|
| EfficientDet | cspdarkdet53 |
| cspdarkdet53m | |
| cspresdet50 | |
| cspresdext50 | |
| cspresdext50pan | |
| efficientdet_d0 | |
| efficientdet_d1 | |
| efficientdet_d5 | |
| efficientdet_em | |
| efficientdet_lite0 | |
| efficientdet_q0 | |
| efficientdet_q1 | |
| efficientdet_q2 | |
| efficientdetv2_ds | |
| efficientdetv2_dt | |
| resdet50 | |
| tf_efficientdet_d0 | |
| tf_efficientdet_d0_ap | |
| tf_efficientdet_d1 | |
| tf_efficientdet_d1_ap | |
| tf_efficientdet_d2 | |
| tf_efficientdet_d2_ap | |
| tf_efficientdet_d3 | |
| tf_efficientdet_d3_ap | |
| tf_efficientdet_d4 | |
| tf_efficientdet_d4_ap | |
| tf_efficientdet_d5 | |
| tf_efficientdet_d5_ap | |
| tf_efficientdet_d6 | |
| tf_efficientdet_lite0 | |
| tf_efficientdet_lite1 | |
| tf_efficientdet_lite2 | |
| tf_efficientdet_lite3 | |
| tf_efficientdet_lite3x | |
| tf_efficientdet_lite4 |
| Model Family | Backbone |
|---|---|
| NanoDet | nanodet-efficient-lite0 |
| nanodet-efficient-lite1 | |
| nanodet-efficient-lite2 | |
| nanodet-m | |
| nanodet-m-1.5x | |
| nanodet-plus-m | |
| nanodet-plus-m-1.5x |
| Model Family | Backbone |
|---|---|
| SSD | mb1-ssd |
| vgg16-ssd |
| Model Family | Backbone |
|---|---|
| YOLOv5 | yolov5l |
| yolov5l6 | |
| yolov5m | |
| yolov5m6 | |
| yolov5n | |
| yolov5n6 | |
| yolov5s | |
| yolov5s6 | |
| yolov5x | |
| yolov5x6 |
| Model Family | Backbone |
|---|---|
| YOLOv8 | yolov8l |
| yolov8m | |
| yolov8n | |
| yolov8s | |
| yolov8x | |
| yolov8x6 |