DBFace models converted from PyTorch to TFJS Graph Model
With native decode bounds and NMS functions in TypeScript to process results
Compatible with
- Browser (using
tfjs-wasm,tfjs-webglortfjs-webgpubackends) - NodeJS (using
tfjs-nodeortfjs-node-gpumodules)
Includes 16 variations of DBFace models:
- MobileNetv2 and MobileNetv3 variations
- Float32 and quantized Float16 variations
- Trained on multiple resolutions: 320x320, 640x480, 960x640 and 1280x800
Examples of outputs of all model variations are included in media folder
## Run
npm start -- --model models/mb3-f32-800-1280/dbface.json --input media/in/models.jpg --output media/out/models-mb3-f32-800x1200.jpg
{ application: 'dbface', version: '0.0.1' }
{
options: {
inImage: 'media/in/models.jpg',
outImage: 'media/out/models-mb3-f32-800x1200.jpg',
modelPath: 'file://models/mb3-f32-800-1280/dbface.json',
minScore: 0.2,
iouThreshold: 0.1,
maxResults: 1000,
inputSize: [ 0, 0 ]
}
}
{ tensorflow: '3.12.0', backend: 'tensorflow', gpuEnabled: true, gpuActive: true }
{ memory: { bytes: 7123912, tensors: 252 } }
{ input: 'media/in/models.jpg', bytes: 24602400, resolution: [ 3600, 2278 ], tensor: [ 1, 800, 1280, 3 ], type: 'float32' }
{ predictTime: 155, processTime: 2, results: 538, nms: 25 }
{ scores: [
0.82, 0.81, 0.81, 0.8, 0.8,
0.8, 0.8, 0.8, 0.79, 0.79,
0.79, 0.79, 0.78, 0.78, 0.78,
0.78, 0.78, 0.77, 0.77, 0.77,
0.75, 0.75, 0.73, 0.72, 0.68
] }
{ output: 'media/out/models-mb3-f32-800x1200.jpg', resolution: [ 3600, 2278 ] }