So sánh phương pháp
Xem các phương pháp đã chọn cạnh nhau; những hàng khác biệt được làm nổi bật.
| Phân loại hình ảnh đa ngôn ngữ× | Phân loại ảnh đa phương thức× | |
|---|---|---|
| Lĩnh vực | Học sâu | Học sâu |
| Họ | Machine learning | Machine learning |
| Năm ra đời≠ | 2020s | 2011–2021 |
| Người khởi xướng≠ | Community / Radford et al. (CLIP, 2021) as key enabler | Ngiam et al.; Radford et al. (CLIP) |
| Loại≠ | Cross-lingual supervised image classification | Multimodal supervised classification |
| Công trình gốc≠ | 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 ↗ | 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 ↗ |
| Tên gọi khác | Cross-lingual image classification, Multilingual visual recognition, Cross-cultural image classification, Multilingual vision-language classification | multimodal visual classification, image-text classification, vision-language classification, cross-modal image classification |
| Liên quan≠ | 5 | 6 |
| Tóm tắt≠ | Multilingual image classification trains visual models to recognise and label images when class names, supervision signals, or evaluation benchmarks span multiple languages. Enabled by multilingual vision-language models such as CLIP, it allows a single model to classify images using prompts or labels in any supported language, facilitating cross-cultural and cross-lingual deployment of computer vision systems. | 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. |
| ScholarGateBộ dữ liệu ↗ |
|
|