목록분류 전체보기 (307)
UOMOP
TopK Activate layer
plot_flops(512, 40);function plot_flops(dim, snr) % 데이터 초기화 full_data = [ 512, 40, 18, 2, 68, 788, 9142; 512, 40, 19, 5, 163, 1428, 8404; 512, 40, 20, 15, 367, 2263, 7355; 512, 40, 21, 49, 758, 3153, 6040; 512, 40, 22, 119, 1396, 3800, 4685; 512, 40, 23, 308, 2206, 4143, 3343; 512, 40, 24, 666, 3123, 3959, 2252; 512, 40, 25, 1281, 387..
CBS activated layer
plot_flops(512, 40);function plot_flops(dim, snr) % 데이터 초기화 full_data = [ 512, 40, 18, 1, 0, 57, 9942; 512, 40, 19, 1, 1, 132, 9866; 512, 40, 20, 3, 4, 297, 9696; 512, 40, 21, 10, 10, 699, 9281; 512, 40, 22, 33, 61, 1287, 8619; 512, 40, 23, 110, 179, 2031, 7680; 512, 40, 24, 321, 396, 2876, 6407; 512, 40, 25, 774, 724, 3478, 5024; ..
X: Target PSNR, Y: FLOPs, SNR : [40, 20], CR: [1/6, 1/12]
% 사용자 입력 받기dim = input('DIM을 입력하세요 (예: 512 또는 256): ');snr = input('SNR을 입력하세요 (예: 40 또는 20): ');% 데이터 정의 (예시로 일부 데이터만 포함)if dim == 512 && snr == 40 target_psnr = 18:30; flops_proposed = [32335.261, 32380.095, 32484.721, 32742.633, 33215.974, 34011.868, 35338.228, 37271.45, 39761.017, 42810.562, 45953.263, 48885.42, 51297.466]; flops_topk = [32865.432, 33389.933, 34213.484, 35417.231, 3..
==================================================================================Proposed==================================================================================dim:512, snr:40, target psnr : 18 Lv1:1, Lv2:0, Lv3:57, Lv4:9942 final FLOPs : 32335.261 Average PSNR over 10000 images: 25.24 dim:512, snr:40, target psnr : 19 Lv1:1, Lv2:1, Lv3:132, Lv4:9866 final FLOPs : 32380.095 Average P..
X : SNR, Y : Avg PSNR
ChessboardDIM:512 SNR(40dB) : (Avg):24.9450 (Ori.):28.88 (MR=33%):28.74 (MR=60%):27.68 (MR=75%):25.201 DIM:512 SNR(30dB) : (Avg):24.5068 (Ori.):27.94 (MR=33%):27.95 (MR=60%):27.24 (MR=75%):24.98DIM:512 SNR(20dB) : (Avg):23.2405 (Ori.):25.83 (MR=33%):25.80 (MR=60%):25.62 (MR=75%):23.89DIM:512 SNR(10dB) : (Avg):21.5922 (Ori.):23.37 (MR=33%):23.36 (MR=60%):22.90 (MR=75%):22.44DIM:512 SNR(0dB) : (..
FLOPs (Proposed encoder)
pip install fvcorepip install thop import torchimport torch.nn as nnfrom thop import profile# 수정된 Encoder 클래스class Encoder(nn.Module): def __init__(self, latent_dim): super(Encoder, self).__init__() self.latent_dim = latent_dim # stride=2를 초반 레이어에 적용하고, kernel_size를 3으로 줄임 self.in1 = nn.Conv2d(3, 32, kernel_size=5, stride=2, padding=3) self.in2 = nn.Conv2d(3..