This page provides performance benchmarks for the DeGirum Orca AI accelerator across various models.
The frames per second (FPS) numbers are obtained by running the single_model_performance_test.ipynb Jupyter Notebook on a machine equipped with ORCA1. The script can also be run in the AI Hub to estimate the performance. All FPS numbers are for batch_size=1. This page will be periodically updated to reflect the latest performance numbers. As our compiler and software mature, we expect to add more models and also improve the performance.
Last updated on: Oct 23, 2023
| Model Name | FPS |
|---|---|
| efficientnet_es_imagenet–224x224_quant | 187 |
| mobiledet_coco–320x320_quant | 128 |
| mobilenet_v1_imagenet–224x224_quant | 407 |
| mobilenet_v2_imagenet–224x224_quant | 360 |
| resnet50_imagenet–224x224_pruned_quant | 250 |
| yolo_v5s_face_det–512x512_quant | 126 |