Initial Release

MTCNN (Face Detection / Landmark Detection) models

All the weights are originally from Iván de Paz Centeno and were converted to tflite format:

• o_net.tflite (sha256: 768385d570300648b7b881acbd418146522b79b4771029bb2e684bdd8c764b9f)
• p_net.tflite (sha 256: 530183192e24f7cc86b6706e1eb600482c4ed4306399ac939c472e3957bae15e)
• r_net.tflite (sha256: 5ec33b065eb2802bc4c2575d21feff1a56958d854785bc3e2907d3b7ace861a2)

Face Recognition models

We trained the following models from scratch using 112x112px aligned MS1M dataset.

• mobileNet.tflite (sha256: 6c19b789f661caa8da735566490bfd8895beffb2a1ec97a56b126f0539991aa6)
• resNet.tflite (sha256: f4d8b0194957a3ad766135505fc70a91343660151a8103bbb6c3b8ac34dbb4e2)
• FaceTransformerOctupletLoss.onnx (sha256: aa995cce8b137ccdc65b394cc57c6b1fdafc7012ce5197e62a4cf8d8e61db4f2)
• MobileNetV2.onnx (sha256: 6f53fb10f0db558403f73cfe744a96b12d763bdf1294a38d14ef14307d61ecf3)
• FaceTransformerOctupletLoss.pt (sha256: b10faa1c170b9fd0f95e3142d9e584ad6f9647d3566207d8bfcc259df8dbdf0f)
• ResNet50.onnx (sha256: 2816d8f4e455525e5f31cd511e1c3f2f677efceefda9fc114e3ac350acc681b7)

Sample Gallery Images:

We provide a small sample gallery containing of 9 images from the LFW dataset.

• sample_gallery.zip (shar256: 99c24edc7edd6dd1a1c306b708c6a17ae5059e446540885e1f1953d2f9125778)