DCT-CompCNN trains CNNs on JPEG compressed DCT coefficients and reports higher accuracy than conventional RGB-input CNNs on Dog vs Cat and CIFAR-10 datasets.
Image classification on CIFAR 10 Dataset,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
DCT-CompCNN: A Novel Image Classification Network Using JPEG Compressed DCT Coefficients
DCT-CompCNN trains CNNs on JPEG compressed DCT coefficients and reports higher accuracy than conventional RGB-input CNNs on Dog vs Cat and CIFAR-10 datasets.