While text-conditional 3D object generation and manipulation have seen rapid progress, the evaluation of coherence between generated 3D shapes and input textual descriptions lacks a clear benchmark. The reason is twofold: a) the low quality of the textual descriptions in the only publicly available dataset of text-shape pairs; b) the limited effectiveness of the metrics used to quantitatively assess such coherence. In this paper, we propose a comprehensive solution that addresses both weaknesses. Firstly, we employ large language models to automatically refine textual descriptions associated with shapes. Secondly, we propose a quantitative metric to assess text-to-shape coherence, through cross-attention mechanisms. To validate our approach, we conduct a user study and compare quantitatively our metric with existing ones. The refined dataset, the new metric and a set of text-shape pairs validated by the user study comprise a novel, fine-grained benchmark that we publicly release to foster research on text-to-shape coherence of text-conditioned 3D generative models. Benchmark available at https://cvlab-unibo.github.io/CrossCoherence-Web/.
Looking at words and points with attention: a benchmark for text-to-shape coherence / Amaduzzi A.; Lisanti G.; Salti S.; Di Stefano L.. - ELETTRONICO. - (2023), pp. 2860-2869. (Intervento presentato al convegno IEEE/CVF International Conference on Computer Vision Workshops tenutosi a Paris, France nel 02-06 October 2023) [10.1109/ICCVW60793.2023.00309].
Looking at words and points with attention: a benchmark for text-to-shape coherence
Amaduzzi A.;Lisanti G.;Salti S.;Di Stefano L.
2023
Abstract
While text-conditional 3D object generation and manipulation have seen rapid progress, the evaluation of coherence between generated 3D shapes and input textual descriptions lacks a clear benchmark. The reason is twofold: a) the low quality of the textual descriptions in the only publicly available dataset of text-shape pairs; b) the limited effectiveness of the metrics used to quantitatively assess such coherence. In this paper, we propose a comprehensive solution that addresses both weaknesses. Firstly, we employ large language models to automatically refine textual descriptions associated with shapes. Secondly, we propose a quantitative metric to assess text-to-shape coherence, through cross-attention mechanisms. To validate our approach, we conduct a user study and compare quantitatively our metric with existing ones. The refined dataset, the new metric and a set of text-shape pairs validated by the user study comprise a novel, fine-grained benchmark that we publicly release to foster research on text-to-shape coherence of text-conditioned 3D generative models. Benchmark available at https://cvlab-unibo.github.io/CrossCoherence-Web/.File | Dimensione | Formato | |
---|---|---|---|
23_ICCVW_Looking_at_Words_and_Points_with_Attention_a_Benchmark_for.pdf
accesso aperto
Tipo:
Postprint
Licenza:
Licenza per accesso libero gratuito
Dimensione
42.3 MB
Formato
Adobe PDF
|
42.3 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.