OPENEDGES Enlight Pro Model Zoo

Enlight Pro Model Zoo

Catalog

Models
87
Tasks
5

Quantization

Weights
INT8 per-channel
Activations
INT8 per-tensor

NPU HW

Compute
8 TOPS per core
MAC array
4K at 1 GHz
Internal memory
2 MB per core

Image Classification

19 models

Network Input Resolution Param Size (M) Source Dataset Accuracy Metric FP32 Original Accuracy Int8 Quantized Accuracy Inference time (ms)
vgg16 224x224 138.36 ONNX Model Zoo ImageNet top1 72.0 71.9 15.86
vgg16-bn 224x224 138.36 ONNX Model Zoo ImageNet top1 72.3 72.3 15.86
vgg19 224x224 143.67 ONNX Model Zoo ImageNet top1 73.3 73.0 17.13
vgg19-bn 224x224 143.67 ONNX Model Zoo ImageNet top1 73.5 73.4 17.13
resnet18 224x224 12.20 ONNX Model Zoo ImageNet top1 69.1 69.1 2.24
resnet34 224x224 22.30 ONNX Model Zoo ImageNet top1 72.7 72.7 3.41
resnet50 224x224 27.58 ONNX Model Zoo ImageNet top1 74.1 73.9 5.66
resnet101 224x224 46.54 ONNX Model Zoo ImageNet top1 75.7 75.4 9.67
resnet152 224x224 62.17 ONNX Model Zoo ImageNet top1 76.0 75.3 13.57
densenet121 224x224 8.09 ONNX Model Zoo ImageNet top1 61.0 57.6 7.63
efficientnet-lite4 380x380 12.95 timm ImageNet top1 81.3 81.2 31.11
squeezenet1.1 224x224 1.24 ONNX Model Zoo ImageNet top1 56.4 56.0 1.52
efficientnet-b0 224x224 6.55 ONNX Model Zoo ImageNet top1 77.7 77.0 9.00
efficientnet-b2 256x256 10.48 ONNX Model Zoo ImageNet top1 79.4 79.0 19.10
efficientnet-b3 288x288 13.73 ONNX Model Zoo ImageNet top1 81.4 80.7 23.15
efficientnet-b4 320x320 21.07 ONNX Model Zoo ImageNet top1 82.6 82.1 34.66
efficientnet-b5 448x448 32.35 ONNX Model Zoo ImageNet top1 85.4 84.8 99.72
mobilenet_v3_large 224x224 5.47 torchvision ImageNet top1 75.3 74.9 5.07
mobilenet_v3_small 224x224 2.54 torchvision ImageNet top1 67.7 66.4 2.78

Object Detection

34 models

Network Input Resolution Param Size (M) Source Dataset Accuracy Metric FP32 Original Accuracy Int8 Quantized Accuracy Inference time (ms)
mb2-ssd-lite 300x300 3.36 oenpu/mb2_ssd_lite_onnx VOC0712 mAP50-95 43.9 43.5 7.88
tf-efficientdet-lite1 384x384 4.40 rwightman/efficientdet-pytorch COCO mAP50-95 28.1 28.1 35.04
tf-efficientdet-lite3 512x512 9.11 rwightman/efficientdet-pytorch COCO mAP50-95 35.4 35.2 80.12
tf-efficientdet-lite4 640x640 16.64 rwightman/efficientdet-pytorch COCO mAP50-95 39.9 39.8 163.10
yolov5n 640x640 1.87 Ultralytics COCO mAP50-95 27.2 26.1 10.61
yolov5s 640x640 7.23 Ultralytics COCO mAP50-95 36.8 35.5 12.76
yolov5m 640x640 21.17 Ultralytics COCO mAP50-95 44.6 43.1 22.99
yolov5l 640x640 46.53 Ultralytics COCO mAP50-95 48.4 45.9 34.55
yolov5x 640x640 86.71 Ultralytics COCO mAP50-95 50.1 48.4 68.58
yolov8n 640x640 3.19 Ultralytics COCO mAP50-95 36.8 36.5 9.80
yolov8s 640x640 11.20 Ultralytics COCO mAP50-95 44.4 44.2 12.58
yolov8m 640x640 25.93 Ultralytics COCO mAP50-95 49.7 49.6 25.27
yolov8l 640x640 43.71 Ultralytics COCO mAP50-95 52.2 52.0 38.41
yolov8x 640x640 68.24 Ultralytics COCO mAP50-95 53.3 53.2 66.98
yolov9t 640x640 2.14 Ultralytics COCO mAP50-95 37.5 37.3 15.58
yolov9s 640x640 7.24 Ultralytics COCO mAP50-95 45.9 45.7 17.76
yolov9m 640x640 20.11 Ultralytics COCO mAP50-95 50.8 50.7 31.15
yolov9c 640x640 25.42 Ultralytics COCO mAP50-95 52.1 52.0 43.40
yolov10n 640x640 2.34 THU-MIG/yolov10 COCO mAP50-95 37.8 37.3 11.66
yolov10s 640x640 7.29 THU-MIG/yolov10 COCO mAP50-95 45.5 45.3 14.43
yolov10m 640x640 15.40 THU-MIG/yolov10 COCO mAP50-95 50.3 47.6 28.75
yolov10b 640x640 19.11 THU-MIG/yolov10 COCO mAP50-95 51.5 47.5 31.82
yolov10l 640x640 24.41 THU-MIG/yolov10 COCO mAP50-95 52.3 47.3 39.56
yolov10x 640x640 29.52 THU-MIG/yolov10 COCO mAP50-95 53.5 50.6 66.32
yolo11n 640x640 2.66 Ultralytics COCO mAP50-95 38.8 38.5 11.38
yolo11s 640x640 9.49 Ultralytics COCO mAP50-95 46.1 45.8 14.20
yolo11m 640x640 20.13 Ultralytics COCO mAP50-95 50.8 50.6 27.17
yolo11l 640x640 25.38 Ultralytics COCO mAP50-95 52.7 52.2 36.24
yolo11x 640x640 56.96 Ultralytics COCO mAP50-95 54.0 51.6 67.26
yolo26n 640x640 2.45 Ultralytics COCO mAP50-95 39.7 38.2 12.71
yolo26s 640x640 9.54 Ultralytics COCO mAP50-95 47.2 45.4 15.95
yolo26m 640x640 20.45 Ultralytics COCO mAP50-95 51.9 50.3 28.12
yolo26l 640x640 24.85 Ultralytics COCO mAP50-95 53.6 52.3 36.92
yolo26x 640x640 55.77 Ultralytics COCO mAP50-95 56.4 54.6 68.23

Instance Segmentation

16 models

Network Input Resolution Param Size (M) Source Dataset Accuracy Metric FP32 Original Accuracy Int8 Quantized Accuracy Inference time (ms)
yolov8n-seg 640x640 3.45 Ultralytics COCO mAP50-95 30.6 30.3 11.05
yolov8s-seg 640x640 11.85 Ultralytics COCO mAP50-95 37.1 37.0 15.07
yolov8m-seg 640x640 27.31 Ultralytics COCO mAP50-95 41.2 41.0 29.67
yolov8l-seg 640x640 46.02 Ultralytics COCO mAP50-95 43.2 43.1 46.23
yolov8x-seg 640x640 71.84 Ultralytics COCO mAP50-95 44.0 43.8 79.11
yolov9c-seg 640x640 27.73 Ultralytics COCO mAP50-95 43.3 42.7 51.78
yolo11n-seg 640x640 2.91 Ultralytics COCO mAP50-95 32.5 32.1 12.64
yolo11s-seg 640x640 10.14 Ultralytics COCO mAP50-95 38.5 38.2 16.69
yolo11m-seg 640x640 22.44 Ultralytics COCO mAP50-95 42.6 42.4 34.98
yolo11l-seg 640x640 27.69 Ultralytics COCO mAP50-95 44.0 43.6 44.05
yolo11x-seg 640x640 62.14 Ultralytics COCO mAP50-95 45.1 42.8 85.12
yolo26n-seg 640x640 2.77 Ultralytics COCO mAP50-95 33.6 32.6 14.53
yolo26s-seg 640x640 10.44 Ultralytics COCO mAP50-95 40.0 39.5 19.65
yolo26m-seg 640x640 23.61 Ultralytics COCO mAP50-95 44.0 43.3 38.83
yolo26l-seg 640x640 28.01 Ultralytics COCO mAP50-95 45.5 44.8 47.63
yolo26x-seg 640x640 62.86 Ultralytics COCO mAP50-95 47.1 46.7 91.14

Semantic Segmentation

3 models

Network Input Resolution Param Size (M) Source Dataset Accuracy Metric FP32 Original Accuracy Int8 Quantized Accuracy Inference time (ms)
pidnet-s 1024x2048 7.62 XuJiacong/PIDNet Cityscapes mIoU 78.8 78.6 89.91
pidnet-m 1024x2048 28.53 XuJiacong/PIDNet Cityscapes mIoU 77.0 77.0 167.97
pidnet-l 1024x2048 36.93 XuJiacong/PIDNet Cityscapes mIoU 80.9 80.7 199.73

Pose Estimation

15 models

Network Input Resolution Param Size (M) Source Dataset Accuracy Metric FP32 Original Accuracy Int8 Quantized Accuracy Inference time (ms)
yolov8n-pose 640x640 3.36 Ultralytics COCO-Pose mAP50-95 50.4 49.1 9.77
yolov8s-pose 640x640 11.68 Ultralytics COCO-Pose mAP50-95 59.7 58.1 13.05
yolov8m-pose 640x640 26.51 Ultralytics COCO-Pose mAP50-95 64.5 63.2 25.89
yolov8l-pose 640x640 44.53 Ultralytics COCO-Pose mAP50-95 67.2 66.1 39.18
yolov8x-pose 640x640 69.53 Ultralytics COCO-Pose mAP50-95 68.7 67.4 68.94
yolo11n-pose 640x640 2.93 Ultralytics COCO-Pose mAP50-95 50.2 48.5 10.61
yolo11s-pose 640x640 9.97 Ultralytics COCO-Pose mAP50-95 58.7 57.6 14.67
yolo11m-pose 640x640 20.96 Ultralytics COCO-Pose mAP50-95 64.8 64.2 27.95
yolo11l-pose 640x640 26.21 Ultralytics COCO-Pose mAP50-95 66.1 65.2 37.02
yolo11x-pose 640x640 58.82 Ultralytics COCO-Pose mAP50-95 69.2 68.7 69.36
yolo26n-pose 640x640 2.99 Ultralytics COCO-Pose mAP50-95 56.7 53.8 12.30
yolo26s-pose 640x640 10.43 Ultralytics COCO-Pose mAP50-95 62.3 60.2 16.88
yolo26m-pose 640x640 21.61 Ultralytics COCO-Pose mAP50-95 68.0 66.0 29.56
yolo26l-pose 640x640 26.00 Ultralytics COCO-Pose mAP50-95 69.5 67.4 38.36
yolo26x-pose 640x640 57.63 Ultralytics COCO-Pose mAP50-95 70.9 69.6 70.33