Image Retrieval on ImageNet-100 (val)
57.04Top-1 AccuracyPlaces
Evaluation results
| Method | Links | |
|---|---|---|
| PlacesPre-training Dataset=Places, Evaluation Protocol=5-NN retrieval, Backbone=ResNet-50, Algorithm=MoCo-v22024.01 | 57.04 | |
| LLMs (w/o GPT-4) + S-21kPre-training Dataset=LLM (w/o GPT-4) combined with Shaders-21k, Evaluation Protocol=5-NN retrieval, Backbone=ResNet-50, Algorithm=MoCo-v22024.01 | 43.4 | |
| LLMs (w/ GPT-4) + S-21kPre-training Dataset=LLM (w/ GPT-4) combined with Shaders-21k, Evaluation Protocol=5-NN retrieval, Backbone=ResNet-50, Algorithm=MoCo-v22024.01 | 43.36 | |
| S-21kPre-training Dataset=S-21k, Evaluation Protocol=5-NN retrieval, Backbone=ResNet-50, Algorithm=MoCo-v22024.01 | 43.24 | |
| StyleGAN OPre-training Dataset=StyleGAN O, Evaluation Protocol=5-NN retrieval, Backbone=ResNet-50, Algorithm=MoCo-v22024.01 | 33 | |
| LLMs (w/ GPT-4)Pre-training Dataset=LLM-generated images (including GPT-4), Evaluation Protocol=5-NN retrieval, Backbone=ResNet-50, Algorithm=MoCo-v22024.01 | 27.44 | |
| LLMs (w/o GPT-4)Pre-training Dataset=LLM-generated images (excluding GPT-4), Evaluation Protocol=5-NN retrieval, Backbone=ResNet-50, Algorithm=MoCo-v22024.01 | 22.42 | |
| FractalDB-1kPre-training Dataset=FractalDB-1k, Evaluation Protocol=5-NN retrieval, Backbone=ResNet-50, Algorithm=MoCo-v22024.01 | 17.24 | |
| Dead-leavesPre-training Dataset=Dead-leaves, Evaluation Protocol=5-NN retrieval, Backbone=ResNet-50, Algorithm=MoCo-v22024.01 | 12.76 | |
| Random InitializationPre-training Dataset=None, Evaluation Protocol=5-NN retrieval, Backbone=ResNet-50, Algorithm=MoCo-v22024.01 | 4.28 |