CCAH: A CLIP-Based Cycle Alignment Hashing Method for Unsupervised Vision-Text Retrieval

Table 1

Comparison results on mean accuracy (MAP@50) for different code lengths under the Flickr-25K and NUS-WIDE dataset.

Dataset	Flickr-25K								NUS-WIDE
Task	I->T				T->I				I->T				T->I
Method	16 bits	32 bits	64 bits	128 bits	16 bits	32 bits	64 bits	128 bits	16 bits	32 bits	64 bits	128 bits	16 bits	32 bits	64 bits	128 bits
CVH	0.606	0.599	0.596	0.598	0.591	0.583	0.576	0.576	0.372	0.362	0.406	0.39	0.401	0.384	0.442	0.432
IMH	0.612	0.601	0.592	0.579	0.603	0.595	0.589	0.58	0.47	0.473	0.476	0.459	0.478	0.483	0.472	0.462
LCMH	0.559	0.569	0.585	0.593	0.561	0.569	0.582	0.582	0.354	0.361	0.389	0.383	0.376	0.387	0.408	0.419
CMFH	0.621	0.624	0.625	0.627	0.642	0.662	0.676	0.685	0.455	0.459	0.465	0.467	0.529	0.577	0.614	0.645
LSSH	0.584	0.599	0.602	0.614	0.618	0.626	0.626	0.628	0.481	0.489	0.507	0.507	0.455	0.459	0.468	0.473
RFDH	0.632	0.636	0.641	0.652	0.681	0.693	0.698	0.702	0.488	0.492	0.494	0.508	0.612	0.641	0.658	0.68
DBRC	0.617	0.619	0.62	0.621	0.618	0.626	0.626	0.628	0.424	0.459	0.447	0.447	0.455	0.459	0.468	0.473
UDCMH	0.689	0.698	0.714	0.717	0.692	0.704	0.718	0.733	0.511	0.519	0.524	0.558	0.637	0.653	0.695	0.716
DJSRH	0.810	0.843	0.862	0.876	0.786	0.822	0.835	0.847	0.724	0.773	0.798	0.817	0.712	0.744	0.771	0.789
DSAH	0.863	0.877	0.895	0.903	0.846	0.860	0.881	0.882	0.775	0.805	0.818	0.827	0.770	0.790	0.804	0.815
HNH	0.853	0.883	0.895	0.902	0.833	0.854	0.868	0.878	0.582	0.747	0.800	0.816	0.423	0.743	0.781	0.780
CCAH	0.863	0.879	0.899	0.910	0.891	0.908	0.914	0.913	0.715	0.754	0.775	0.787	0.834	0.851	0.864	0.874