List your company
ResearchFormerly WizWand

Image Classification on ImageNet 1k (val) (Accuracy & Latency)

84.41Top-1 AccuracyViT-Base
69.111673.083377.05581.0267Mar 27, 2024
Updated 8mo ago

Evaluation results

MethodLinks
ViT-Basecondition=Original2024.0384.41111.1
DC-ViT-BMACs=15.5 x 10^9, Number of compressed blocks (k)=1, Number of training samples=10002024.0383.78101.9
DC-ViT-BMACs=15.5 x 10^9, Number of compressed blocks (k)=1, Number of training samples=5002024.0383.73101.9
DC-ViT-BMACs=15.5 x 10^9, Number of compressed blocks (k)=1, Number of training samples=1002024.0383.3101.9
DC-ViT-BMACs=15.5 x 10^9, Number of compressed blocks (k)=1, Number of training samples=502024.0382.88101.9
DC-ViT-BMACs=14.1 x 10^9, Number of compressed blocks (k)=2, Number of training samples=502024.0380.5993.5
DC-ViT-BMACs=12.7 x 10^9, Number of compressed blocks (k)=3, Number of training samples=5002024.0380.1884.9
PRACTISEMACs=15.5 x 10^9, Number of compressed blocks (k)=1, Number of training samples=100, implemented_by_us=true2024.0375.69103.2
PRACTISEMACs=15.5 x 10^9, Number of compressed blocks (k)=1, Number of training samples=1000, implemented_by_us=true2024.0374.71103.2
PRACTISEMACs=15.5 x 10^9, Number of compressed blocks (k)=1, Number of training samples=500, implemented_by_us=true2024.0373.32103.2
PRACTISEMACs=15.5 x 10^9, Number of compressed blocks (k)=1, Number of training samples=50, implemented_by_us=true2024.0371.28103.2
PRACTISEMACs=12.7 x 10^9, Number of compressed blocks (k)=3, Number of training samples=500, Training epochs=60002024.0370.5485.9
PRACTISEMACs=12.7 x 10^9, Number of compressed blocks (k)=3, Number of training samples=500, implemented_by_us=true2024.0370.0385.9
PRACTISEMACs=14.1 x 10^9, Number of compressed blocks (k)=2, Number of training samples=50, Training epochs=40002024.0369.8794.7
PRACTISEMACs=14.1 x 10^9, Number of compressed blocks (k)=2, Number of training samples=50, implemented_by_us=true2024.0369.794.7