retinanet-examples评估COCO 2017

https://github.com/NVIDIA/retinanet-examples

Evaluate your detection model on COCO 2017:

root@72066437f2c8:/workspace/data/model/retinanet_rn50fpn# odtk infer retinanet_rn50fpn.pth --images /workspace/data/val2017/ --annotations /workspace/data/instances_val2017.json
Loading model from retinanet_rn50fpn.pth...
     model: RetinaNet
  backbone: ResNet50FPN
   classes: 80, anchors: 9
Preparing dataset...
    loader: pytorch
    resize: 800, max: 1333
   backend: pytorch
    device: 1 GPU
     batch: 2, precision: mixed
 BBOX type: axis aligned
Running inference...
[  90/5000] 1.353s/2-batch (fw: 1.350s), 1.5 im/s
[ 178/5000] 1.385s/2-batch (fw: 1.385s), 1.4 im/s
[ 266/5000] 1.371s/2-batch (fw: 1.371s), 1.5 im/s
[ 360/5000] 1.288s/2-batch (fw: 1.288s), 1.6 im/s
[ 448/5000] 1.365s/2-batch (fw: 1.365s), 1.5 im/s
[ 540/5000] 1.324s/2-batch (fw: 1.324s), 1.5 im/s
[ 628/5000] 1.389s/2-batch (fw: 1.388s), 1.4 im/s
[ 716/5000] 1.381s/2-batch (fw: 1.381s), 1.4 im/s
[ 806/5000] 1.344s/2-batch (fw: 1.344s), 1.5 im/s
[ 890/5000] 1.436s/2-batch (fw: 1.435s), 1.4 im/s
[ 982/5000] 1.320s/2-batch (fw: 1.319s), 1.5 im/s
[1072/5000] 1.339s/2-batch (fw: 1.339s), 1.5 im/s
[1162/5000] 1.354s/2-batch (fw: 1.354s), 1.5 im/s
[1252/5000] 1.345s/2-batch (fw: 1.345s), 1.5 im/s
[1342/5000] 1.361s/2-batch (fw: 1.360s), 1.5 im/s
[1436/5000] 1.299s/2-batch (fw: 1.299s), 1.5 im/s
[1526/5000] 1.352s/2-batch (fw: 1.351s), 1.5 im/s
[1616/5000] 1.358s/2-batch (fw: 1.357s), 1.5 im/s
[1708/5000] 1.307s/2-batch (fw: 1.307s), 1.5 im/s
[1800/5000] 1.344s/2-batch (fw: 1.344s), 1.5 im/s
[1890/5000] 1.371s/2-batch (fw: 1.371s), 1.5 im/s
[1980/5000] 1.335s/2-batch (fw: 1.334s), 1.5 im/s
[2066/5000] 1.414s/2-batch (fw: 1.413s), 1.4 im/s
[2154/5000] 1.383s/2-batch (fw: 1.383s), 1.4 im/s
[2246/5000] 1.327s/2-batch (fw: 1.327s), 1.5 im/s
[2338/5000] 1.328s/2-batch (fw: 1.327s), 1.5 im/s
[2428/5000] 1.336s/2-batch (fw: 1.336s), 1.5 im/s
[2518/5000] 1.346s/2-batch (fw: 1.346s), 1.5 im/s
[2606/5000] 1.382s/2-batch (fw: 1.382s), 1.4 im/s
[2694/5000] 1.375s/2-batch (fw: 1.374s), 1.5 im/s
[2784/5000] 1.357s/2-batch (fw: 1.357s), 1.5 im/s
[2872/5000] 1.367s/2-batch (fw: 1.367s), 1.5 im/s
[2960/5000] 1.379s/2-batch (fw: 1.379s), 1.5 im/s
[3050/5000] 1.358s/2-batch (fw: 1.357s), 1.5 im/s
[3140/5000] 1.350s/2-batch (fw: 1.349s), 1.5 im/s
[3228/5000] 1.387s/2-batch (fw: 1.387s), 1.4 im/s
[3316/5000] 1.383s/2-batch (fw: 1.382s), 1.4 im/s
[3410/5000] 1.302s/2-batch (fw: 1.302s), 1.5 im/s
[3500/5000] 1.346s/2-batch (fw: 1.346s), 1.5 im/s
[3590/5000] 1.365s/2-batch (fw: 1.364s), 1.5 im/s
[3678/5000] 1.387s/2-batch (fw: 1.387s), 1.4 im/s
[3766/5000] 1.377s/2-batch (fw: 1.377s), 1.5 im/s
[3852/5000] 1.413s/2-batch (fw: 1.412s), 1.4 im/s
[3940/5000] 1.376s/2-batch (fw: 1.376s), 1.5 im/s
[4028/5000] 1.376s/2-batch (fw: 1.375s), 1.5 im/s
[4118/5000] 1.341s/2-batch (fw: 1.341s), 1.5 im/s
[4206/5000] 1.370s/2-batch (fw: 1.369s), 1.5 im/s
[4298/5000] 1.314s/2-batch (fw: 1.313s), 1.5 im/s
[4384/5000] 1.407s/2-batch (fw: 1.407s), 1.4 im/s
[4478/5000] 1.282s/2-batch (fw: 1.281s), 1.6 im/s
[4566/5000] 1.389s/2-batch (fw: 1.388s), 1.4 im/s
[4656/5000] 1.347s/2-batch (fw: 1.346s), 1.5 im/s
[4746/5000] 1.338s/2-batch (fw: 1.338s), 1.5 im/s
[4834/5000] 1.368s/2-batch (fw: 1.368s), 1.5 im/s
[4922/5000] 1.391s/2-batch (fw: 1.391s), 1.4 im/s
[5000/5000] 1.337s/2-batch (fw: 1.337s), 1.5 im/s
Gathering results...
Writing detections.json...
Evaluating model...
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.00001
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.00001
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.00000
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.00000
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.00000
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.00001
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.00035
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.00136
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.00136
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.00000
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.00000
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.00354