Benchmarks
The Hailo-8™ AI Processor brings industry-leading neural processing throughput and power efficiency to support a wide range of AI applications. The table below demonstrates some of Hailo-8’s best-in-class capabilities in the foundational neural tasks of object detection and classification. It achieves very high FPS and power efficiency (FPS/W) across a number of industry-standard neural network models, while also offloading postprocessing from the host processor its working with.
Benchmarks
NN Model | Input Resolution | FPS | Power [W] | FPS/ W |
---|---|---|---|---|
Classification | ||||
ResNet-50 v1 | 224×224 | 1,331 | 3.3 | 402 |
MobileNet_v2_1.0 | 224×224 | 2,444 | 2.162 | 1,130 |
EfficientNet_M | 240×240 | 890 | 3.4 | 261 |
Object Detection | ||||
SSD_MobileNet_v1 | 300×300 | 1,055 | 2.3 | 458 |
YOLOv5m | 640×640 | 218 | 4.4 | 49.5 |
Segmentation | ||||
stdc1 | 1024×1920 | 34 | 2.6 | 13 |
Multi stream object detection (8 streams) | ||||
YOLOv3 | 608×608 | 66 | 4.4 | 15 |
Notes:
- Measurements were taken at room temperature through PCIe interface on Hailo-8 evaluation board
- System host: Intel® Core™ i5-9400 CPU @ 2.90GHz; Models compiled with Hailo Dataflow Compiler version 3.23.0 (SW version 2023-04)
Click here and here to see how the Hailo AI Processor compares.