Face recognition l: ONNX iResNet100

Click a block to read its description, or select it with Tab and Enter.

Face recognition l: ONNX iResNet100Aligned RGB input 112 × 112, batch 1, 512-dimensional embedding. Actual exported ONNX operators, with fused convolution parameters.LibreYOLOFace recognition l: ONNX iResNet100Aligned RGB input 112 × 112, batch 1, 512-dimensional embedding. Actual exported ONNX operators, with fused convolution parameters.Recognition networkAligned RGB face3 × 112 × 112Normalize (x - 127.5) / 127.53 × 112 × 112Conv 3×3 / 13 to 64, p=1; fused biasPReLU64 × 112 × 112Layer 164 × 56 × 56, n=3Layer 2128 × 28 × 28, n=13Layer 3256 × 14 × 14, n=30Layer 4512 × 7 × 7, n=3BatchNormalization512 × 7 × 7Flatten25,088Gemm25,088 to 512, bias=TrueBatchNormalization512L2 normalize512-dimensional identity vectorLayer 1Downsample block, n=1Input 64 × 112 × 112BatchNormalization64 × 112 × 112Conv 3×3 / 164 × 112 × 112, p=1PRelu64 × 112 × 112Conv 3×3 / 264 × 56 × 56, p=1+Conv 1×1 / 264 × 56 × 56Output 64 × 56 × 56Identity block, n=2Input 64 × 56 × 56BatchNormalization64 × 56 × 56Conv 3×3 / 164 × 56 × 56, p=1PRelu64 × 56 × 56Conv 3×3 / 164 × 56 × 56, p=1+Output 64 × 56 × 56Layer 2Downsample block, n=1Input 64 × 56 × 56BatchNormalization64 × 56 × 56Conv 3×3 / 1128 × 56 × 56, p=1PRelu128 × 56 × 56Conv 3×3 / 2128 × 28 × 28, p=1+Conv 1×1 / 2128 × 28 × 28Output 128 × 28 × 28Identity block, n=12Input 128 × 28 × 28BatchNormalization128 × 28 × 28Conv 3×3 / 1128 × 28 × 28, p=1PRelu128 × 28 × 28Conv 3×3 / 1128 × 28 × 28, p=1+Output 128 × 28 × 28Layer 3Downsample block, n=1Input 128 × 28 × 28BatchNormalization128 × 28 × 28Conv 3×3 / 1256 × 28 × 28, p=1PRelu256 × 28 × 28Conv 3×3 / 2256 × 14 × 14, p=1+Conv 1×1 / 2256 × 14 × 14Output 256 × 14 × 14Identity block, n=29Input 256 × 14 × 14BatchNormalization256 × 14 × 14Conv 3×3 / 1256 × 14 × 14, p=1PRelu256 × 14 × 14Conv 3×3 / 1256 × 14 × 14, p=1+Output 256 × 14 × 14Layer 4Downsample block, n=1Input 256 × 14 × 14BatchNormalization256 × 14 × 14Conv 3×3 / 1512 × 14 × 14, p=1PRelu512 × 14 × 14Conv 3×3 / 2512 × 7 × 7, p=1+Conv 1×1 / 2512 × 7 × 7Output 512 × 7 × 7Identity block, n=2Input 512 × 7 × 7BatchNormalization512 × 7 × 7Conv 3×3 / 1512 × 7 × 7, p=1PRelu512 × 7 × 7Conv 3×3 / 1512 × 7 × 7, p=1+Output 512 × 7 × 7Artifact verificationlibrefacerec-l.onnx: 103 Conv, 51 BatchNormalization, 50 PRelu, 49 Add, one Flatten and one Gemm.SHA-256 a7933ea5330113b01c9b60351d8f4c33003f145d8470ac5f0e52ee2effe25c60. Shape inference pinned batch 1.Source: libreyolo/models/facerec/model.py. Revision a4d0ecc9e17f.libreyolo.com