ProCAP: Probabilistic Cross-Attentive Prompt Learning for Vision-Language Models

Benchmark Model Rank Results
prompt-engineering-on-oxford-iiit-pet-datasetProCAP#2Harmonic mean: 97.34
prompt-engineering-on-ucf101ProCAP#2Harmonic mean: 83.91
prompt-engineering-on-caltech-101ProCAP#3Harmonic mean: 97.11
prompt-engineering-on-eurosatProCAP#3Harmonic mean: 90.98
prompt-engineering-on-food-101ProCAP#5Harmonic mean: 91.34
prompt-engineering-on-stanford-carsProCAP#7Harmonic mean: 77.95
prompt-engineering-on-imagenet-v2ProCAP#8Top-1 accuracy %: 64.60
prompt-engineering-on-fgvc-aircraftProCAP#9Harmonic mean: 40.84
prompt-engineering-on-imagenet-sProCAP#9Top-1 accuracy %: 49.53
prompt-engineering-on-dtdProCAP#11Harmonic mean: 73.17
prompt-engineering-on-imagenet-aProCAP#12Top-1 accuracy %: 50.83
prompt-engineering-on-sun397ProCAP#12Harmonic mean: 80.66
prompt-engineering-on-imagenet-rProCAP#13Top-1 accuracy %: 77.35
prompt-engineering-on-oxford-102-flowerProCAP#15Harmonic mean: 85.50
prompt-engineering-on-imagenetProCAP#19Harmonic mean: 73.48