FER2013 quantize result

gistfile1.txt

Benchmark model: mobilenet_small
Benchmark for non quantized model

LOAD DATASET: START
LOAD DATASET: DONE
INFERENCE: START
INFERENCE: DONE

Angry: Correct cases: 301 / 491, Accuracy: 61.30
Disgust: Correct cases: 0 / 55, Accuracy: 0.00
Fear: Correct cases: 176 / 528, Accuracy: 33.33
Happy: Correct cases: 728 / 879, Accuracy: 82.82
Sad: Correct cases: 242 / 594, Accuracy: 40.74
Surprise: Correct cases: 270 / 416, Accuracy: 64.90
Neutral: Correct cases: 397 / 626, Accuracy: 63.42
All category: Correct cases: 2114 / 3589, Accuracy: 58.90

0.58 ms for one data

Benchmark for quantized model
LOAD DATASET: START
LOAD DATASET: DONE
INFERENCE: START
INFERENCE: DONE

Angry: Correct cases: 304 / 491, Accuracy: 61.91
Disgust: Correct cases: 0 / 55, Accuracy: 0.00
Fear: Correct cases: 181 / 528, Accuracy: 34.28
Happy: Correct cases: 727 / 879, Accuracy: 82.71
Sad: Correct cases: 241 / 594, Accuracy: 40.57
Surprise: Correct cases: 271 / 416, Accuracy: 65.14
Neutral: Correct cases: 384 / 626, Accuracy: 61.34
All category: Correct cases: 2108 / 3589, Accuracy: 58.74

18.46 ms for one data