List your company

Image-to-audio Retrieval on MIAO (test)

61.45mAPProposed method
4.582819.346434.1148.8736May 17, 2026

Evaluation results

MethodLinks
Proposed methodDirection=I2A, Image Encoder=CLIP (frozen), Audio Encoder=CLAP (frozen), Projector=Two-layer MLP2026.0561.4553.668.960
BaselineDirection=I2A, Image Encoder=CLIP (ViT-B/32), Audio Encoder=CLAP (HTS-AT)2026.056.77297.6