Bandingkan kaedah
Semak kaedah pilihan anda secara bersebelahan; baris yang berbeza akan diserlahkan.
| Klasifikasi Imej Multimodal× | Penyematan Zarah Pelbagai Mod (Multimodal Sentence Embeddings)× | |
|---|---|---|
| Bidang | Pembelajaran Mendalam | Pembelajaran Mendalam |
| Keluarga | Machine learning | Machine learning |
| Tahun asal≠ | 2011–2021 | 2013–2021 |
| Pengasas≠ | Ngiam et al.; Radford et al. (CLIP) | Frome et al. (DeViSE, 2013); popularized by Radford et al. (CLIP, 2021) |
| Jenis≠ | Multimodal supervised classification | Representation learning model |
| Sumber perintis≠ | Radford, A., Kim, J. W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., ... & Sutskever, I. (2021). Learning transferable visual models from natural language supervision. Proceedings of the 38th International Conference on Machine Learning (ICML), PMLR 139, 8748–8763. link ↗ | Radford, A., Kim, J. W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., ... & Sutskever, I. (2021). Learning transferable visual models from natural language supervision. In Proceedings of the 38th International Conference on Machine Learning (ICML), pp. 8748–8763. PMLR. link ↗ |
| Alias | multimodal visual classification, image-text classification, vision-language classification, cross-modal image classification | multimodal embeddings, cross-modal sentence embeddings, vision-language embeddings, joint image-text embeddings |
| Berkaitan≠ | 6 | 1 |
| Ringkasan≠ | Multimodal image classification extends standard visual classification by incorporating additional modalities — such as text captions, audio, or structured metadata — alongside image features. Separate encoders process each modality, their representations are fused, and a joint classifier assigns the target label. Models such as CLIP demonstrate that image–text alignment enables zero-shot and few-shot image classification at scale. | Multimodal sentence embeddings map text and images (and sometimes audio or video) into a shared continuous vector space, so that semantically related pairs from different modalities land close together. Trained by contrastive objectives on large paired corpora, these representations power cross-modal retrieval, zero-shot classification, and vision-language reasoning. |
| ScholarGateSet data ↗ |
|
|