Semantic Image Transmission on ImageNet 256x256 (test)
0.122LPIPSDiT-JSCC
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DiT-JSCCSNR=5dB, Channel=AWGN, CBR=1/482026.01 | 0.122 | 0.135 | 0.97 | 0.064 | 0.969 | 49 | |
| HiFi-DiffComSNR=5dB, Channel=AWGN, CBR=1/48, Variant=NTSCC2026.01 | 0.137 | 0.14 | 0.961 | 0.074 | 0.947 | 57 | |
| HiFi-DiffComSNR=5dB, Channel=AWGN, CBR=1/48, Variant=Deep JSCC2026.01 | 0.165 | 0.164 | 0.959 | 0.089 | 0.938 | 69 | |
| DiT-JSCCSNR=0dB, Channel=AWGN, CBR=1/482026.01 | 0.166 | 0.151 | 0.963 | 0.097 | 0.958 | 60 | |
| SwinJSCCSNR=5dB, Channel=AWGN, CBR=1/48, Training SNR=7dB2026.01 | 0.188 | 0.194 | 0.939 | 0.104 | 0.937 | 95 | |
| HiFi-DiffComSNR=0dB, Channel=AWGN, CBR=1/48, Variant=NTSCC2026.01 | 0.211 | 0.191 | 0.936 | 0.143 | 0.9 | 107 | |
| SwinJSCCSNR=0dB, Channel=AWGN, CBR=1/48, Training SNR=1dB2026.01 | 0.261 | 0.233 | 0.915 | 0.178 | 0.878 | 138 | |
| HiFi-DiffComSNR=0dB, Channel=AWGN, CBR=1/48, Variant=Deep JSCC2026.01 | 0.283 | 0.203 | 0.933 | 0.192 | 0.88 | 115 | |
| VTM + LDPCSNR=5dB, Channel=AWGN, CBR=1/482026.01 | 0.321 | 0.267 | 0.881 | 0.231 | 0.797 | 198 | |
| BPG + LDPCSNR=5dB, Channel=AWGN, CBR=1/482026.01 | 0.351 | 0.272 | 0.874 | 0.271 | 0.808 | 197 | |
| VTM + LDPCSNR=0dB, Channel=AWGN, CBR=1/482026.01 | 0.427 | 0.313 | 0.844 | 0.36 | 0.72 | 244 | |
| BPG + LDPCSNR=0dB, Channel=AWGN, CBR=1/482026.01 | 0.444 | 0.314 | 0.828 | 0.402 | 0.717 | 252 |