Research Article
An FPGA-Based Convolutional Neural Network Coprocessor
Table 2
The performance statistics of the coprocessor computing VGG16.
| | Layer | Number of operations (Ops) | Number of clocks (clk_num) | Performance (GOP/s) | Power (W) |
| | Model | 9437184 | 21669 | 87.10 | 9.21 | | Conv1 | 176518656 | 471803 | 74.83 | 9.21 | | Conv2 | 3765731328 | 33739520 | 22.32 | 9.23 | | Conv3 | 1916338176 | 4287402 | 89.39 | 9.24 | | Conv4 | 3832676352 | 17419712 | 44.00 | 9.23 | | Conv5 | 1984167936 | 2440995 | 162.57 | 9.22 | | Conv6 | 3968335872 | 9996224 | 79.40 | 9.23 | | Conv7 | 3968335872 | 9996224 | 79.40 | 9.25 | | Conv8 | 2123366400 | 3468295 | 122.44 | 9.28 | | Conv9 | 4246732800 | 7790023 | 109.03 | 9.25 | | Conv10 | 4246732800 | 7790023 | 109.03 | 9.26 | | Conv11 | 1207959552 | 3621895 | 66.70 | 9.23 | | Conv12 | 1207959552 | 3621895 | 66.70 | 9.24 | | Conv13 | 1207959552 | 3621895 | 66.70 | 9.23 |
|
|
