Table 9 Hit Ratio of each script identifier in Benchmark 3 - ResNet Model. The best performances for each task and training option are highlighted in bold
From: MDIW-13: a New Multi-Lingual and Multi-Script Database and Benchmark for Script Identification
Train with | Test with | |||||
---|---|---|---|---|---|---|
Handwritten | Printed | |||||
Docs | Lines | Words | Docs | Lines | Words | |
Task 1 | ||||||
Handwritten docs | 78.43% | 47.54% | 39.49% | 27.14% | 28.96% | 27.87% |
Handwritten lines | 47.41% | 89.92% | 71.68% | 33.02% | 30.09% | 29.35 |
Handwritten words | 47.38% | 87.91% | 87.72% | 42.17% | 48.63% | 43.75% |
Printed docs | 18.19% | 29.01% | 25.55% | 93.55% | 95.64% | 85.16% |
Printed Lines | 19.88% | 30.82% | 28.08% | 91.28% | 99.53% | 95.67% |
Printed words | 18.02% | 29.67% | 30.56% | 90.28% | 96.81% | 96.46% |
Task 2 | ||||||
Handwritten docs, lines and words | 86.24% | 92.48% | 89.14% | 52.58% | 52.02% | 45.61% |
Printed docs, lines and words | 21.49% | 36.87% | 30.93% | 96.84% | 99.82% | 98.57% |
Task 3 | ||||||
Printed and Handwritten docs, lines and words | 29.26% | 40.04% | 34.20% | 96.48% | 99.78% | 99.09% |