Skip to content

Latest commit

 

History

History
237 lines (132 loc) · 8.16 KB

File metadata and controls

237 lines (132 loc) · 8.16 KB

Public Models

All models were compiled using Hailo Dataflow Compiler v2.18.0.


Text_Image_Retrieval


Link Legend


Key / Icon Description
Networks used by Hailo-apps.
S Source – Link to the model's open-source repository.
PT Pretrained – Download the pretrained model file (ZIP format).
HEF, NV12, RGBX Compiled Models – Links to models in various formats: - HEF: RGB format - NV12: NV12 format - RGBX: RGBX format
PR Profiler Report – Download the model's performance profiling report.


Network Name float Retrieval@10 Hardware Retrieval@10 FPS (Batch Size=1) FPS (Batch Size=8) Links Input Resolution (HxWxC) Params (M) OPS (G)
clip_resnet_50_text_encoder 88.8 83.9 27.7 84.1 1x77x512 37.8 6.0
clip_resnet_50x4_text_encoder 91.2 89.9 17.9 65.3 1x77x640 59.1 9.3
clip_vit_b_16_text_encoder 90.9 89.8 26.0 68.2 1x77x512 37.8 6.0
clip_vit_b_32_text_encoder⭐ 90.6 89.3 30.1 96.4 1x77x512 37.8 6.0
siglip2_b_32_256_text_encoder 96.1 96.2 12.9 35.5 8x8x768 85.6 11.0
siglip_b_16_text_encoder 96.2 96.0 13.5 36.3 8x8x768 85.6 11.1
tinyclip_vit_39m_16_text_19m_yfcc15m_text_encoder 94.0 94.0 55.5 194 1x77x512 19 3
tinyclip_vit_40m_32_text_19m_laion400m_text_encoder 91.1 90.0 59.2 175 1x77x512 19 3
tinyclip_vit_61m_32_text_29m_laion400m_text_encoder 93.8 90.0 39.1 123 1x77x512 29 4.5
tinyclip_vit_8m_16_text_3m_yfcc15m_text_encoder 84.4 84.2 257 975 1x77x512 3 382