Face Identification

The models provide 16 x float32 or 8 x float32 output vectors.
Data distribution allows float32 packing to int16 values, providing 32-byte or 16-byte respective int8 vectors.

Terminology

Face Detection - finding face boundaries on the image.
Face Recognition (Identification) - finding whether two face photos belong to the same person.

Hardware

Graphcore IPUs x1/x4/x16
Nvidia GPUs x4/x16/x32 (torch: DataParallel, DistributedDataParallel)
Google TPUs (no meaningful results)

Credits

Google Colaboratory (fast networking)
Kaggle (datasets)
spell.ml (workflows, Graphcore hardware)
Paperspace (workflows, Graphcore hardware)
immers.cloud (dozens of GPUs, unlimited RAM, fast SSDs)

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
gc		gc
ic		ic
FaceId.ipynb		FaceId.ipynb
FaceId_2080.ipynb		FaceId_2080.ipynb
FaceId_2080_2.ipynb		FaceId_2080_2.ipynb
FaceId_2080_3.ipynb		FaceId_2080_3.ipynb
FaceId_2080_bs=gc.ipynb		FaceId_2080_bs=gc.ipynb
FaceId_2080_sm.ipynb		FaceId_2080_sm.ipynb
FaceId_2xa100_cudnn.ipynb		FaceId_2xa100_cudnn.ipynb
FaceId_3x2080.ipynb		FaceId_3x2080.ipynb
FaceId_DataIter.ipynb		FaceId_DataIter.ipynb
FaceId_DataIter_v2.ipynb		FaceId_DataIter_v2.ipynb
FaceId_DataIter_v3.ipynb		FaceId_DataIter_v3.ipynb
FaceId_DataIter_v4.ipynb		FaceId_DataIter_v4.ipynb
FaceId_DataIter_v4_2.ipynb		FaceId_DataIter_v4_2.ipynb
FaceId_DataSave.ipynb		FaceId_DataSave.ipynb
FaceId_Normalization.ipynb		FaceId_Normalization.ipynb
FaceId_ResnetGray.ipynb		FaceId_ResnetGray.ipynb
FaceId_a100.ipynb		FaceId_a100.ipynb
FaceId_a100_cudnn.ipynb		FaceId_a100_cudnn.ipynb
FaceId_a100_cudnn_2.ipynb		FaceId_a100_cudnn_2.ipynb
FaceId_convert.ipynb		FaceId_convert.ipynb
FaceId_gc.ipynb		FaceId_gc.ipynb
FaceId_gc_bench.ipynb		FaceId_gc_bench.ipynb
FaceId_gc_bs8_ga16.ipynb		FaceId_gc_bs8_ga16.ipynb
FaceId_gc_ds=nv.ipynb		FaceId_gc_ds=nv.ipynb
FaceId_gc_ov8.ipynb		FaceId_gc_ov8.ipynb
FaceId_gc_ov8_bak.ipynb		FaceId_gc_ov8_bak.ipynb
FaceId_gc_test_bs.ipynb		FaceId_gc_test_bs.ipynb
FaceId_gc_test_bs_full.ipynb		FaceId_gc_test_bs_full.ipynb
FaceId_gc_test_bs_sw.ipynb		FaceId_gc_test_bs_sw.ipynb
FaceId_gc_test_bs_sw_full.ipynb		FaceId_gc_test_bs_sw_full.ipynb
FaceId_gc_test_x16.ipynb		FaceId_gc_test_x16.ipynb
FaceId_gc_test_x2.ipynb		FaceId_gc_test_x2.ipynb
FaceId_gc_test_x4.ipynb		FaceId_gc_test_x4.ipynb
FaceId_gc_test_x4_err.ipynb		FaceId_gc_test_x4_err.ipynb
FaceId_gc_test_x4_err2.ipynb		FaceId_gc_test_x4_err2.ipynb
FaceId_gpu.ipynb		FaceId_gpu.ipynb
FaceId_infer.ipynb		FaceId_infer.ipynb
FaceId_infer_old.ipynb		FaceId_infer_old.ipynb
FaceId_infer_ov8.ipynb		FaceId_infer_ov8.ipynb
FaceId_infer_ov8_sl.ipynb		FaceId_infer_ov8_sl.ipynb
FaceId_infer_sl.ipynb		FaceId_infer_sl.ipynb
FaceId_k.ipynb		FaceId_k.ipynb
FaceId_k_16.ipynb		FaceId_k_16.ipynb
FaceId_k_8.ipynb		FaceId_k_8.ipynb
FaceId_multigpu.ipynb		FaceId_multigpu.ipynb
FaceId_multigpu_bench.ipynb		FaceId_multigpu_bench.ipynb
FaceId_multigpu_bench2_4xa100.ipynb		FaceId_multigpu_bench2_4xa100.ipynb
FaceId_multigpu_bench_4xa100.ipynb		FaceId_multigpu_bench_4xa100.ipynb
FaceId_multigpu_ddp_test.ipynb		FaceId_multigpu_ddp_test.ipynb
FaceId_multigpu_distributed.ipynb		FaceId_multigpu_distributed.ipynb
FaceId_multigpu_distributed_ZeRO.ipynb		FaceId_multigpu_distributed_ZeRO.ipynb
FaceId_multigpu_distributed_bbf.ipynb		FaceId_multigpu_distributed_bbf.ipynb
FaceId_multigpu_distributed_err.ipynb		FaceId_multigpu_distributed_err.ipynb
FaceId_multigpu_distributed_err_x1.ipynb		FaceId_multigpu_distributed_err_x1.ipynb
FaceId_multigpu_distributed_x1.ipynb		FaceId_multigpu_distributed_x1.ipynb
FaceId_multigpu_distributed_x2-0.ipynb		FaceId_multigpu_distributed_x2-0.ipynb
FaceId_multigpu_distributed_x2-1.ipynb		FaceId_multigpu_distributed_x2-1.ipynb
FaceId_multigpu_distributed_x4-0.ipynb		FaceId_multigpu_distributed_x4-0.ipynb
FaceId_multigpu_distributed_x4-1.ipynb		FaceId_multigpu_distributed_x4-1.ipynb
FaceId_multigpu_distributed_x4-2.ipynb		FaceId_multigpu_distributed_x4-2.ipynb
FaceId_multigpu_distributed_x4-3.ipynb		FaceId_multigpu_distributed_x4-3.ipynb
FaceId_multigpu_distributed_x4_autograd_detect_anomaly.ipynb		FaceId_multigpu_distributed_x4_autograd_detect_anomaly.ipynb
FaceId_multigpu_distributed_x8-0.ipynb		FaceId_multigpu_distributed_x8-0.ipynb
FaceId_multigpu_distributed_x8-1.ipynb		FaceId_multigpu_distributed_x8-1.ipynb
FaceId_multigpu_distributed_x8-2.ipynb		FaceId_multigpu_distributed_x8-2.ipynb
FaceId_multigpu_distributed_x8-3.ipynb		FaceId_multigpu_distributed_x8-3.ipynb
FaceId_multigpu_distributed_x8-4.ipynb		FaceId_multigpu_distributed_x8-4.ipynb
FaceId_multigpu_distributed_x8-5.ipynb		FaceId_multigpu_distributed_x8-5.ipynb
FaceId_multigpu_distributed_x8-6.ipynb		FaceId_multigpu_distributed_x8-6.ipynb
FaceId_multigpu_distributed_x8-7.ipynb		FaceId_multigpu_distributed_x8-7.ipynb
FaceId_multigpu_distributed_x8_v2.ipynb		FaceId_multigpu_distributed_x8_v2.ipynb
FaceId_multigpu_x1_sm.ipynb		FaceId_multigpu_x1_sm.ipynb
FaceId_multigpu_x2_sm.ipynb		FaceId_multigpu_x2_sm.ipynb
FaceId_multigpu_x4_maxbatch_sm.ipynb		FaceId_multigpu_x4_maxbatch_sm.ipynb
FaceId_multigpu_x4_sm.ipynb		FaceId_multigpu_x4_sm.ipynb
FaceId_t4.ipynb		FaceId_t4.ipynb
FaceId_t4_sm.ipynb		FaceId_t4_sm.ipynb
FaceId_tpu_dp.ipynb		FaceId_tpu_dp.ipynb
FaceId_tpu_dp_k.ipynb		FaceId_tpu_dp_k.ipynb
FaceId_tpu_pl.ipynb		FaceId_tpu_pl.ipynb
FaceId_v100.ipynb		FaceId_v100.ipynb
FaceId_v100_sm.ipynb		FaceId_v100_sm.ipynb
Parallel_test.ipynb		Parallel_test.ipynb
Pretrained_ResNet_with_grayscale_images_k.ipynb		Pretrained_ResNet_with_grayscale_images_k.ipynb
Pretrained_ResNet_with_grayscale_images_k_cpu.ipynb		Pretrained_ResNet_with_grayscale_images_k_cpu.ipynb
Pretrained_ResNet_with_grayscale_images_v1.ipynb		Pretrained_ResNet_with_grayscale_images_v1.ipynb
Pretrained_ResNet_with_grayscale_images_v2.ipynb		Pretrained_ResNet_with_grayscale_images_v2.ipynb
Pretrained_ResNet_with_grayscale_images_v3.ipynb		Pretrained_ResNet_with_grayscale_images_v3.ipynb
PyTorch_XLA.ipynb		PyTorch_XLA.ipynb
TaskQueue.ipynb		TaskQueue.ipynb
faceid-model-0.011796065969605762.pt		faceid-model-0.011796065969605762.pt
faceid-model-0.016750329740267147.pt		faceid-model-0.016750329740267147.pt
faceid-model-0.02253507034381968.pt		faceid-model-0.02253507034381968.pt
faceid-model-ov8-0.016422454035053595.pt		faceid-model-ov8-0.016422454035053595.pt
faceid-model-ov8-0.02309628876459473.pt		faceid-model-ov8-0.02309628876459473.pt
faceid-preprocess.ipynb		faceid-preprocess.ipynb
facenet_infer.ipynb		facenet_infer.ipynb
facenet_infer_coljit.ipynb		facenet_infer_coljit.ipynb

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

Face Identification

Terminology

Hardware

Credits

TL;DR

About

Uh oh!

Contributors

Uh oh!

Languages

Folders and files

Latest commit

History

Repository files navigation

Face Identification

Terminology

Hardware

Credits

TL;DR

About

Resources

Uh oh!

Stars

Watchers

Forks

Contributors

Uh oh!

Languages