Object Detection on PASCAL VOC x8 scale 2012 (test)
36.7mAPIHR → T
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| IHR → TPipeline Variant=IHR → T, SR Method=Oracle (IHR), Task Network=Faster-RCNN, Backbone=MobileNetV32024.04 | 36.7 | 0 | — | |
| SR4IR (SwinIR)Pipeline Variant=Ours, SR Method=SwinIR, Task Network=Faster-RCNN, Backbone=MobileNetV32024.04 | 27.8 | 0.382 | 21.07 | |
| SR4IR (EDSR)Pipeline Variant=Ours, SR Method=EDSR-baseline, Task Network=Faster-RCNN, Backbone=MobileNetV32024.04 | 25.5 | 0.416 | 20.86 | |
| S → T (SwinIR)Pipeline Variant=S → T, SR Method=SwinIR, Task Network=Faster-RCNN, Backbone=MobileNetV32024.04 | 23.7 | 0.474 | 21.85 | |
| S → T (EDSR)Pipeline Variant=S → T, SR Method=EDSR-baseline, Task Network=Faster-RCNN, Backbone=MobileNetV32024.04 | 21.9 | 0.494 | 21.62 | |
| S + T (SwinIR)Pipeline Variant=S + T, SR Method=SwinIR, Task Network=Faster-RCNN, Backbone=MobileNetV32024.04 | 21.9 | 0.489 | 21.11 | |
| S + T (EDSR)Pipeline Variant=S + T, SR Method=EDSR-baseline, Task Network=Faster-RCNN, Backbone=MobileNetV32024.04 | 20.3 | 0.506 | 20.98 | |
| ILR → TPipeline Variant=ILR → T, SR Method=Bilinear up, Task Network=Faster-RCNN, Backbone=MobileNetV32024.04 | 18.9 | 0.559 | 20.29 | |
| T → S (SwinIR)Pipeline Variant=T → S, SR Method=SwinIR, Task Network=Faster-RCNN, Backbone=MobileNetV32024.04 | 18.9 | 0.449 | 20.59 | |
| T → S (EDSR)Pipeline Variant=T → S, SR Method=EDSR-baseline, Task Network=Faster-RCNN, Backbone=MobileNetV32024.04 | 15.5 | 0.476 | 20.48 |