Multi-label Image Classification on BigEarthNet
63.6AccuracyVGG19
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| VGG19Training Time (hours.mins)=0.57, Inference Rate (imgs/sec)=218, Model Size=20,038,1632021.11 | 63.6 | 81.7 | 74.2 | 77.7 | |
| VGG16Training Time (hours.mins)=0.49, Inference Rate (imgs/sec)=242, Model Size=14,728,4672021.11 | 63 | 81.4 | 73.6 | 77.3 | |
| ViTM/20Training Time (hours.mins)=0.29, Inference Rate (imgs/sec)=586, Model Size=9,286,4192021.11 | 62.7 | 80.6 | 73.8 | 77.1 | |
| ResNet50Framework=Distributed Learning, Training Time (hours.mins)=0.59, Inference Rate (imgs/sec)=351, Model Size=23,648,5952021.11 | 62.4 | 78.8 | 75 | 76.8 | |
| DenseNet121Training Time (hours.mins)=0.50, Inference Rate (imgs/sec)=407, Model Size=7,078,9312021.11 | 62.3 | 79.2 | 74.5 | 76.8 | |
| ViT/20Training Time (hours.mins)=0.23, Inference Rate (imgs/sec)=720, Model Size=7,760,1472021.11 | 62.1 | 80.5 | 73.1 | 76.6 | |
| ResNet50Framework=Single GPU, Training Time (hours.mins)=13.22, Inference Rate (imgs/sec)=345, Model Size=23,648,5952021.11 | 61.8 | 78.1 | 74.8 | 76.4 | |
| ResNet101Training Time (hours.mins)=1.38, Inference Rate (imgs/sec)=268, Model Size=42,719,0592021.11 | 61.7 | 78.8 | 74 | 76.3 | |
| WRNB0-ECATraining Time (hours.mins)=0.11, Inference Rate (imgs/sec)=823, Model Size=306,8172021.11 | 61.7 | 81.5 | 71.8 | 76.3 | |
| ViT/30Training Time (hours.mins)=0.20, Inference Rate (imgs/sec)=704, Model Size=5,458,7072021.11 | 61.6 | 80.1 | 72.7 | 76.2 | |
| EfficientNetB0-CBAMTraining Time (hours.mins)=0.17, Inference Rate (imgs/sec)=501, Model Size=4,412,8192021.11 | 61.5 | 79.9 | 72.7 | 76.1 | |
| WRNB0-SETraining Time (hours.mins)=0.11, Inference Rate (imgs/sec)=751, Model Size=309,7292021.11 | 61.5 | 81.1 | 71.8 | 76.2 | |
| EfficientNetB0-SETraining Time (hours.mins)=0.15, Inference Rate (imgs/sec)=640, Model Size=4,411,2512021.11 | 61.4 | 79.6 | 72.8 | 76.1 | |
| EfficientNetB0-ECATraining Time (hours.mins)=0.14, Inference Rate (imgs/sec)=651, Model Size=3,461,6632021.11 | 61.4 | 79.6 | 72.9 | 76.1 | |
| EfficientNetB0-ECA-GHOSTTraining Time (hours.mins)=0.15, Inference Rate (imgs/sec)=611, Model Size=2,103,6632021.11 | 61.4 | 80.4 | 72.1 | 76 | |
| ViT/12Training Time (hours.mins)=0.36, Inference Rate (imgs/sec)=668, Model Size=15,984,9152021.11 | 61.3 | 80.7 | 71.9 | 76 | |
| EfficientNetB0-COORD-GHOSTTraining Time (hours.mins)=0.19, Inference Rate (imgs/sec)=509, Model Size=2,833,9672021.11 | 61.3 | 80 | 72.4 | 76 | |
| EfficientNetB0-COORDTraining Time (hours.mins)=0.18, Inference Rate (imgs/sec)=604, Model Size=4,191,9672021.11 | 61.2 | 79.3 | 72.8 | 75.9 | |
| DenseNet169Training Time (hours.mins)=1.01, Inference Rate (imgs/sec)=370, Model Size=12,696,4672021.11 | 61.1 | 77.7 | 74.1 | 75.9 | |
| EfficientNetB0-CBAM-GHOSTTraining Time (hours.mins)=0.18, Inference Rate (imgs/sec)=471, Model Size=3,054,8192021.11 | 61 | 80.5 | 71.7 | 75.8 | |
| ResNet152Training Time (hours.mins)=2.19, Inference Rate (imgs/sec)=203, Model Size=58,431,8272021.11 | 60.8 | 76.2 | 75 | 75.6 | |
| DenseNet201Training Time (hours.mins)=1.17, Inference Rate (imgs/sec)=325, Model Size=18,380,4352021.11 | 60.8 | 78.1 | 73.3 | 75.6 | |
| EfficientNetB0-SE-GHOSTTraining Time (hours.mins)=0.16, Inference Rate (imgs/sec)=602, Model Size=3,053,2512021.11 | 60.7 | 80.2 | 71.4 | 75.5 | |
| WRNB0-CBAMTraining Time (hours.mins)=0.13, Inference Rate (imgs/sec)=639, Model Size=310,0232021.11 | 60.6 | 80.2 | 71.3 | 75.5 | |
| ViT/40Training Time (hours.mins)=0.19, Inference Rate (imgs/sec)=691, Model Size=4,989,2032021.11 | 60.4 | 80.1 | 71 | 75.3 | |
| MLPMixerTraining Time (hours.mins)=0.12, Inference Rate (imgs/sec)=807, Model Size=446,7232021.11 | 60.3 | 79.1 | 71.8 | 75.2 | |
| WRNB0-CBAM-GHOSTTraining Time (hours.mins)=0.12, Inference Rate (imgs/sec)=670, Model Size=160,8392021.11 | 59.9 | 80 | 70.5 | 74.9 | |
| WRNB0-COORDTraining Time (hours.mins)=0.18, Inference Rate (imgs/sec)=588, Model Size=312,7472021.11 | 59.8 | 80.9 | 69.6 | 74.8 | |
| WRNB0-ECA-GHOSTTraining Time (hours.mins)=0.12, Inference Rate (imgs/sec)=851, Model Size=157,6332021.11 | 59.7 | 80.7 | 69.7 | 74.8 | |
| WRNB0-SE-GHOSTTraining Time (hours.mins)=0.12, Inference Rate (imgs/sec)=808, Model Size=160,5452021.11 | 59.6 | 80.8 | 69.5 | 74.7 | |
| ResNet50-GHOSTTraining Time (hours.mins)=1.04, Inference Rate (imgs/sec)=403, Model Size=11,930,3872021.11 | 58.4 | 75.2 | 72.4 | 73.8 | |
| WRNB0-COORD-GHOSTTraining Time (hours.mins)=0.19, Inference Rate (imgs/sec)=655, Model Size=163,5632021.11 | 58.3 | 80.4 | 68 | 73.7 | |
| MLPMixer/20Training Time (hours.mins)=0.10, Inference Rate (imgs/sec)=780, Model Size=740,2552021.11 | 58 | 78.1 | 69.4 | 73.4 | |
| MLPMixer/6Training Time (hours.mins)=0.17, Inference Rate (imgs/sec)=654, Model Size=468,0832021.11 | 57.8 | 76.1 | 70.6 | 73.2 | |
| WRNB0Training Time (hours.mins)=0.10, Inference Rate (imgs/sec)=807, Model Size=306,8032021.11 | 56.5 | 80 | 65.9 | 72.2 | |
| MLPMixerTinyTraining Time (hours.mins)=0.08, Inference Rate (imgs/sec)=911, Model Size=40,8632021.11 | 55.7 | 77.5 | 66.5 | 71.6 | |
| WRNB0-GHOSTTraining Time (hours.mins)=0.11, Inference Rate (imgs/sec)=845, Model Size=157,6192021.11 | 54.9 | 79.5 | 64 | 70.9 | |
| ViT/6Training Time (hours.mins)=2.33, Inference Rate (imgs/sec)=241, Model Size=55,237,3952021.11 | 54.1 | 76.8 | 64.6 | 70.2 | |
| MLPMixer/30Training Time (hours.mins)=0.11, Inference Rate (imgs/sec)=845, Model Size=1,369,7152021.11 | 53.5 | 77.1 | 63.6 | 69.7 | |
| MLPMixer/40Training Time (hours.mins)=0.12, Inference Rate (imgs/sec)=823, Model Size=2,261,9912021.11 | 49.6 | 75.1 | 59.4 | 66.3 | |
| K-BranchTraining Time (hours.mins)=1.01, Inference Rate (imgs/sec)=510, Model Size=36,979,0272021.11 | 47.3 | 63.4 | 65.1 | 64.2 |