Complex Handwriting Trajectory Recovery: Evaluation Metrics and Algorithm

AbstractMany important tasks such as forensic signature verification, calligraphy synthesis, etc., rely on handwriting trajectory recovery of which, however, even an appropriate evaluation metric is still missing. Indeed, existing metrics only focus the writing orders but overlook fidelity glyphs. Taking both facets into account, we come up with two new metrics, adaptive intersection union (AIoU) which eliminates influence various stroke widths, and length-independent dynamic time warping (LDTW) solves trajectory-point alignment problem. After that, then propose a novel model named Parsing-and-tracing ENcoder-decoder Network (PEN-Net), in particular for characters complex glyph long trajectory, was believed very challenging. In PEN-Net, carefully designed double-stream parsing encoder parses structure, global tracing decoder overcomes memory difficulty prediction. Our experiments demonstrate that AIoU LDTW together can truly assess quality proposed PEN-Net exhibits satisfactory performance complex-glyph languages including Chinese, Japanese Indic. The source code available at https://github.com/ChenZhounan/PEN-Net.KeywordsTrajectory recoveryHandwritingEvaluation