List your company
ResearchFormerly WizWand

Image Classification on ICinW 20 datasets 1.0 (test)

63.39AccuracyK-LITE
25.700435.485245.2755.0548Apr 20, 2022
Updated 8mo ago

Evaluation results

MethodLinks
K-LITETraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=28M (full), Evaluation Protocol=Fine-tuning, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0463.39
UniCLTraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=28M (full), Evaluation Protocol=Fine-tuning, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0460.99
K-LITETraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=15M (half), Evaluation Protocol=Fine-tuning, Few-shot setting=5-shot, Data Fusion Scheme=Combine2022.0460.72
UniCLTraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=15M (half), Evaluation Protocol=Fine-tuning, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0459.04
K-LITETraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=28M (full), Evaluation Protocol=Linear Probing, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0458.56
UniCLTraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=27M (full), Evaluation Protocol=Fine-tuning, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0458.25
K-LITETraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=27M (full), Evaluation Protocol=Fine-tuning, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0458.24
K-LITETraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=15M (half), Evaluation Protocol=Fine-tuning, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0458.2
UniCLTraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=28M (full), Evaluation Protocol=Linear Probing, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0457.92
K-LITETraining Dataset=ImageNet-21K, Number of Training Samples=13M (full), Evaluation Protocol=Fine-tuning, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0457.81
K-LITETraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=15M (half), Evaluation Protocol=Linear Probing, Few-shot setting=5-shot, Data Fusion Scheme=Combine2022.0457.38
K-LITETraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=27M (full), Evaluation Protocol=Linear Probing, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0457.06
UniCLTraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=14M (half), Evaluation Protocol=Fine-tuning, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0456.45
UniCLTraining Dataset=ImageNet-21K, Number of Training Samples=13M (full), Evaluation Protocol=Fine-tuning, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0455.96
UniCLTraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=27M (full), Evaluation Protocol=Linear Probing, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0455.96
K-LITETraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=15M (half), Evaluation Protocol=Linear Probing, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0455.91
K-LITETraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=14M (half), Evaluation Protocol=Fine-tuning, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0455.88
K-LITETraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=14M (half), Evaluation Protocol=Linear Probing, Few-shot setting=5-shot, Data Fusion Scheme=Combine2022.0454.28
K-LITETraining Dataset=ImageNet-21K, Number of Training Samples=13M (full), Evaluation Protocol=Linear Probing, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0453.92
UniCLTraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=15M (half), Evaluation Protocol=Linear Probing, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0453.86
UniCLTraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=14M (half), Evaluation Protocol=Linear Probing, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0453.5
UniCLTraining Dataset=ImageNet-21K, Number of Training Samples=13M (full), Evaluation Protocol=Linear Probing, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0453.07
K-LITETraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=14M (half), Evaluation Protocol=Fine-tuning, Few-shot setting=5-shot, Data Fusion Scheme=Combine2022.0452.11
K-LITETraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=14M (half), Evaluation Protocol=Linear Probing, Few-shot setting=5-shot, Data Fusion Scheme=Concat2022.0449.48
K-LITETraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=28M (full), Evaluation Protocol=Zero-shot, Data Fusion Scheme=Concat2022.0441.34
K-LITETraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=15M (half), Evaluation Protocol=Zero-shot, Data Fusion Scheme=Combine2022.0440.32
K-LITETraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=15M (half), Evaluation Protocol=Zero-shot, Data Fusion Scheme=Concat2022.0439.53
UniCLTraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=28M (full), Evaluation Protocol=Zero-shot, Data Fusion Scheme=Concat2022.0438.9
K-LITETraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=27M (full), Evaluation Protocol=Zero-shot, Data Fusion Scheme=Concat2022.0438.89
K-LITETraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=14M (half), Evaluation Protocol=Zero-shot, Data Fusion Scheme=Concat2022.0436.5
K-LITETraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=14M (half), Evaluation Protocol=Zero-shot, Data Fusion Scheme=Combine2022.0436.5
UniCLTraining Dataset=GCC-15M + ImageNet-21K, Number of Training Samples=15M (half), Evaluation Protocol=Zero-shot, Data Fusion Scheme=Concat2022.0436.31
UniCLTraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=27M (full), Evaluation Protocol=Zero-shot, Data Fusion Scheme=Concat2022.0435.99
UniCLTraining Dataset=YFCC-14M + ImageNet-21K, Number of Training Samples=14M (half), Evaluation Protocol=Zero-shot, Data Fusion Scheme=Concat2022.0434.3
K-LITETraining Dataset=ImageNet-21K, Number of Training Samples=13M (full), Evaluation Protocol=Zero-shot, Data Fusion Scheme=Concat2022.0433.44
UniCLTraining Dataset=ImageNet-21K, Number of Training Samples=13M (full), Evaluation Protocol=Zero-shot, Data Fusion Scheme=Concat2022.0427.15