Zero-Shot Composed Image Retrieval with Textual Inversion

Benchmark Model Rank Results
zero-shot-composed-image-retrieval-zs-cir-onSEARLE-XL (CLIP L/14)#32mAP@10: 12.73
zero-shot-composed-image-retrieval-zs-cir-onSEARLE (CLIP B/32)#37mAP@10: 9.94
zero-shot-composed-image-retrieval-zs-cir-on-1SEARLE#40R@5: 53.42
zero-shot-composed-image-retrieval-zs-cir-on-1SEARLE-XL#42R@5: 52.48
zero-shot-composed-image-retrieval-zs-cir-on-11SEARLE (CLIP B/32)#4A-R@1: 14.4
zero-shot-composed-image-retrieval-zs-cir-on-11SEARLE (CLIP L/14)#5A-R@1: 14.4
zero-shot-composed-image-retrieval-zs-cir-on-2SEARLE-XL-OTI (CLIP L/14)#30(Recall@10+Recall@50)/2: 37.76
zero-shot-composed-image-retrieval-zs-cir-on-2SEARLE-XL (CLIP L/14)#33(Recall@10+Recall@50)/2: 35.90
zero-shot-composed-image-retrieval-zs-cir-on-2SEARLE (CLIP B/32)#37(Recall@10+Recall@50)/2: 32.71
zero-shot-composed-image-retrieval-zs-cir-on-2SEARLE-OTI (CLIP B/32)#38(Recall@10+Recall@50)/2: 32.39
zero-shot-composed-image-retrieval-zs-cir-on-4SEARLE-XL-OTI (CLIP L/14)#2Actions Recall@5: 31.43
zero-shot-composed-image-retrieval-zs-cir-on-4SEARLE-XL (CLIP L/14)#4Actions Recall@5: 29.02
zero-shot-composed-image-retrieval-zs-cir-on-4SEARLE-OTI (CLIP B/32)#8Actions Recall@5: 26.00
zero-shot-composed-image-retrieval-zs-cir-on-4SEARLE (CLIP B/32)#10Actions Recall@5: 24.58
zero-shot-composed-image-retrieval-zs-cir-on-5SEARLE-XL (CLIP L/14)#4Average Recall: 21.54
zero-shot-composed-image-retrieval-zs-cir-on-5SEARLE-XL-OTI (CLIP B/32)#5Average Recall: 20.42
zero-shot-composed-image-retrieval-zs-cir-on-5SEARLE-OTI (CLIP B/32)#10Average Recall: 12.77
zero-shot-composed-image-retrieval-zs-cir-on-5SEARLE (CLIP B/32)#11Average Recall: 11.94
zero-shot-composed-image-retrieval-zs-cir-on-6SEARLE-XL (CLIP L/14)#7(Recall@10+Recall@50)/2: 21.54
zero-shot-composed-image-retrieval-zs-cir-on-6SEARLE-XL-OTI (CLIP B/32)#8(Recall@10+Recall@50)/2: 20.42
zero-shot-composed-image-retrieval-zs-cir-on-6SEARLE-OTI (CLIP B/32)#13(Recall@10+Recall@50)/2: 12.77
zero-shot-composed-image-retrieval-zs-cir-on-6SEARLE (CLIP B/32)#14(Recall@10+Recall@50)/2: 11.94