Model card with bench
Browse files
README.md
ADDED
|
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Yolov7-pose Benchmark results
|
| 2 |
+
|
| 3 |
+
Benchmark, comparing CPU/iGPU with 3 different processors:
|
| 4 |
+
|
| 5 |
+
- i5-1250pe (alder lake)
|
| 6 |
+
- i5-1145G7E (tiger lake)
|
| 7 |
+
- i7-1185G7E (tiger lake)
|
| 8 |
+
|
| 9 |
+
|
| 10 |
+
3 different models:
|
| 11 |
+
- FP16
|
| 12 |
+
- Partial QAT (model is quantized but the head is preserved for accuracy)
|
| 13 |
+
- int8 QAT (model is fully quantized)
|
| 14 |
+
|
| 15 |
+
|Device - Model |Precision |Average Latency (ms) |FPS|
|
| 16 |
+
| --- | --- | --- | --- |
|
| 17 |
+
|i5-1250pe-CPU |int8 QAT |11.63 |85.10|
|
| 18 |
+
|i7-1185G7E CPU |int8 QAT |13.36 |74.50|
|
| 19 |
+
|i7-1185G7E iGPU |Partial QAT |13.51 |73.32|
|
| 20 |
+
|i7-1185G7E iGPU |int8 QAT |13.66 |72.53|
|
| 21 |
+
|i5-1250pe-CPU |Partial QAT |14.14 |68.75|
|
| 22 |
+
|i5-1145G7E CPU |int8 QAT |16.32 |61.01|
|
| 23 |
+
|i7-1185G7E CPU |Partial QAT |17.30 |60.39|
|
| 24 |
+
|i7-1185G7E iGPU |FP16 |17.52 |56.63|
|
| 25 |
+
|i5-1145G7E iGPU |Partial QAT |17.91 |55.78|
|
| 26 |
+
|i5-1145G7E iGPU |int8 QAT |17.95 |55.66|
|
| 27 |
+
|i5-1250pe-iGPU |Partial QAT |18.99 |51.04|
|
| 28 |
+
|i5-1250pe-iGPU |int8 QAT |20.12 |49.29|
|
| 29 |
+
|i5-1145G7E CPU |Partial QAT |20.28 |49.14|
|
| 30 |
+
|i5-1250pe-iGPU |FP16 |21.25 |46.71|
|
| 31 |
+
|i5-1145G7E iGPU |FP16 |23.86 |42.73|
|
| 32 |
+
|i7-1185G7E CPU |FP16 |37.03 |26.96|
|
| 33 |
+
|i5-1250pe-CPU |FP16 |37.83 |26.38|
|
| 34 |
+
|i5-1145G7E CPU |FP16 |44.27 |22.56|
|