Table 4 Performance comparison of four HUNet_24 variants on the MTACCR test_1 set

From: HUNet: hierarchical universal network for multi-type ancient Chinese character recognition

Model

Top-1 (%)

Params (M)

Flops (M)

MemR + W (MB)

Input

Feature_dim

Total depths

Throughput (images/s)

GPU

CPU

ONNX

HUNet_24(Ours)

92.59

11.82 (1.43)

261

74.59

128

1280

12 (2,2,6,2)

17,103

50.8

108.1

HUNet_24_all_ECA

92.51

11.82 (1.44)

259

65.02

128

1280

12 (2,2,6,2)

14,148

48.2

71.3

HUNet_24_no_ECA

91.96

11.81 (1.43)

258

65.02

128

1280

12 (2,2,6,2)

14,235

52.1

76.4

HUNet_24_local_ECA

92.55

11.81 (1.43)

258

65.02

128

1280

12 (2,2,6,2)

13,742

50.6

75.6

HUNet_24_global_ECA

92.15

11.81 (1.43)

258

65.02

128

1280

12 (2,2,6,2)

13,680

50.8

75.7

  1. Boldface indicates the best performance in the test set, while underlined text denotes the second-best results.