Skip to content

Latest commit

 

History

History
503 lines (491 loc) · 19.5 KB

HAILO15H_classification.rst

File metadata and controls

503 lines (491 loc) · 19.5 KB

Public Pre-Trained Models

Here, we give the full list of publicly pre-trained models supported by the Hailo Model Zoo.

  • Benchmark Networks are marked with rocket
  • Networks available in TAPPAS are marked with star
  • Benchmark and TAPPAS networks run in performance mode
  • All models were compiled using Hailo Dataflow Compiler v3.29.0

ImageNet

Network Name Accuracy (top1) HW Accuracy FPS (Batch Size=1) FPS (Batch Size=8) Input Resolution (HxWxC) Params (M) OPS (G) Pretrained Source Compiled
cas_vit_m 81.2 81.02 42 91 384x384x3 12.42 10.89 download link rgbx
cas_vit_s 79.93 79.75 59 131 384x384x3 5.5 5.4 download link rgbx
cas_vit_t 81.9 81.56 31 63 384x384x3 21.76 20.85 download link rgbx
deit_base 80.93 80.1 39 107 224x224x3 80.26 35.22 download link rgbx
deit_small 76.25 75.51 47 115 224x224x3 20.52 9.4 download link rgbx
deit_tiny 69.07 68.45 97 375 224x224x3 5.3 2.57 download link rgbx
efficientnet_l 80.47 79.3 106 222 300x300x3 10.55 19.4 download link rgbx
efficientnet_lite0 74.99 73.85 1009 1008 224x224x3 4.63 0.78 download link rgbx
efficientnet_lite1 76.67 76.3 275 776 240x240x3 5.39 1.22 download link rgbx
efficientnet_lite2 77.46 76.73 182 476 260x260x3 6.06 1.74 download link rgbx
efficientnet_lite3 79.29 78.66 143 358 280x280x3 8.16 2.8 download link rgbx
efficientnet_lite4 80.79 80.08 103 255 300x300x3 12.95 5.10 download link rgbx
efficientnet_m 78.91 78.48 171 410 240x240x3 6.87 7.32 download link rgbx
efficientnet_s 77.63 77.3 341 341 224x224x3 5.41 4.72 download link rgbx
fastvit_sa12 76.8 76.74 188 616 224x224x3 11.99 3.59 download link rgbx
hardnet39ds 73.43 73.01 380 1208 224x224x3 3.48 0.86 download link rgbx
hardnet68 75.47 75.24 157 370 224x224x3 17.56 8.5 download link rgbx
inception_v1 69.74 69.59 355 842 224x224x3 6.62 3 download link rgbx
mobilenet_v1 70.97 70.3 2873 2873 224x224x3 4.22 1.14 download link rgbx
mobilenet_v2_1.0 rocket 71.78 70.85 3454 3454 224x224x3 3.49 0.62 download link rgbx
mobilenet_v2_1.4 74.18 73.18 579 579 224x224x3 6.09 1.18 download link rgbx
mobilenet_v3 72.21 71.81 374 1100 224x224x3 4.07 2 download link rgbx
mobilenet_v3_large_minimalistic 72.12 70.6 2596 2596 224x224x3 3.91 0.42 download link rgbx
regnetx_1.6gf 77.05 76.71 376 1080 224x224x3 9.17 3.22 download link rgbx
regnetx_800mf 75.16 74.82 2558 2558 224x224x3 7.24 1.6 download link rgbx
repghost_1_0x 73.03 72.33 235 702 224x224x3 4.1 0.28 download link rgbx
repghost_2_0x 77.18 76.9 150 461 224x224x3 9.8 1.04 download link rgbx
repvgg_a1 74.4 72.46 1782 1782 224x224x3 12.79 4.7 download link rgbx
repvgg_a2 76.52 74.42 257 529 224x224x3 25.5 10.2 download link rgbx
resmlp12_relu 75.27 74.84 87 308 224x224x3 15.77 6.04 download link rgbx
resnet_v1_18 71.27 70.77 2030 2030 224x224x3 11.68 3.64 download link rgbx
resnet_v1_34 72.7 72.23 268 694 224x224x3 21.79 7.34 download link rgbx
resnet_v1_50 rocket star 75.21 74.6 290 969 224x224x3 25.53 6.98 download link rgbx
resnext26_32x4d 76.17 75.96 352 822 224x224x3 15.37 4.96 download link rgbx
resnext50_32x4d 79.3 78.46 199 511 224x224x3 24.99 8.48 download link rgbx
squeezenet_v1.1 59.85 59.34 2775 2775 224x224x3 1.24 0.78 download link rgbx
vit_base 84.5 83.15 39 107 224x224x3 86.5 35.188 download link rgbx
vit_base_bn rocket 79.98 79.27 64 202 224x224x3 86.5 35.188 download link rgbx
vit_small 81.5 80.01 70 250 224x224x3 21.12 8.62 download link rgbx
vit_small_bn 78.12 77.09 127 460 224x224x3 21.12 8.62 download link rgbx
vit_tiny 75.51 73.89 96 370 224x224x3 5.73 2.2 download link rgbx
vit_tiny_bn 68.95 67.14 229 1068 224x224x3 5.73 2.2 download link rgbx