Hi, I see your answer about calculating FPS is see s/img printed in terminal (or saved to log file) when inference is performed. Inverting that will give FPS, but I don't see s/img in the log file I get, only s/iter. Below I give the result of completing one round. Can s/iter be regarded as the FPS calculation?
[06/23 17:17:15 d2.engine.hooks]: Overall training speed: 179998 iterations in 1 day, 4:10:13 (0.5634 s / it)
[06/23 17:17:15 d2.engine.hooks]: Total training time: 1 day, 5:06:54 (0:56:41 on hooks)
INFO:croptrain.data.datasets.visdrone:Loaded 548 images in COCO format from /root/siton-data-zhangmingData/YHJ/detectron2/croptrain/datasets/VisDrone/VisDrone2019-DET_val_coco.json
[06/23 17:17:17 d2.data.common]: Serializing the dataset using: <class 'detectron2.data.common._TorchSerializedList'>
[06/23 17:17:17 d2.data.common]: Serializing 548 elements to byte tensors and concatenating them all ...
[06/23 17:17:17 d2.data.common]: Serialized dataset takes 1.45 MiB
[06/23 17:17:17 d2.data.dataset_mapper]: [DatasetMapper] Augmentations used in inference: [ResizeShortestEdge(short_edge_length=(1200, 1200), max_size=1999, sample_style='choice')]
INFO:croptrain.engine.inference:Start inference on 548 batches
INFO:croptrain.engine.inference:Inference done 11/548. Dataloading: 0.0016 s/iter. Inference: 0.0860 s/iter. Eval: 0.0009 s/iter. Total: 0.0885 s/iter. ETA=0:00:47
INFO:croptrain.engine.inference:Inference done 74/548. Dataloading: 0.0020 s/iter. Inference: 0.0776 s/iter. Eval: 0.0008 s/iter. Total: 0.0805 s/iter. ETA=0:00:38
INFO:croptrain.engine.inference:Inference done 132/548. Dataloading: 0.0021 s/iter. Inference: 0.0781 s/iter. Eval: 0.0032 s/iter. Total: 0.0834 s/iter. ETA=0:00:34
INFO:croptrain.engine.inference:Inference done 190/548. Dataloading: 0.0021 s/iter. Inference: 0.0784 s/iter. Eval: 0.0039 s/iter. Total: 0.0845 s/iter. ETA=0:00:30
INFO:croptrain.engine.inference:Inference done 253/548. Dataloading: 0.0021 s/iter. Inference: 0.0781 s/iter. Eval: 0.0031 s/iter. Total: 0.0834 s/iter. ETA=0:00:24
INFO:croptrain.engine.inference:Inference done 309/548. Dataloading: 0.0021 s/iter. Inference: 0.0787 s/iter. Eval: 0.0038 s/iter. Total: 0.0846 s/iter. ETA=0:00:20
INFO:croptrain.engine.inference:Inference done 370/548. Dataloading: 0.0021 s/iter. Inference: 0.0788 s/iter. Eval: 0.0033 s/iter. Total: 0.0842 s/iter. ETA=0:00:14
INFO:croptrain.engine.inference:Inference done 427/548. Dataloading: 0.0021 s/iter. Inference: 0.0788 s/iter. Eval: 0.0038 s/iter. Total: 0.0847 s/iter. ETA=0:00:10
INFO:croptrain.engine.inference:Inference done 490/548. Dataloading: 0.0021 s/iter. Inference: 0.0785 s/iter. Eval: 0.0034 s/iter. Total: 0.0841 s/iter. ETA=0:00:04
INFO:croptrain.engine.inference:Total inference time: 0:00:45.689610 (0.084143 s / iter per device, on 1 devices)
INFO:croptrain.engine.inference:Total inference pure compute time: 0:00:42 (0.078699 s / iter per device, on 1 devices)
Hi, I see your answer about calculating FPS is see s/img printed in terminal (or saved to log file) when inference is performed. Inverting that will give FPS, but I don't see s/img in the log file I get, only s/iter. Below I give the result of completing one round. Can s/iter be regarded as the FPS calculation?
[06/23 17:17:15 d2.engine.hooks]: Overall training speed: 179998 iterations in 1 day, 4:10:13 (0.5634 s / it) [06/23 17:17:15 d2.engine.hooks]: Total training time: 1 day, 5:06:54 (0:56:41 on hooks) INFO:croptrain.data.datasets.visdrone:Loaded 548 images in COCO format from /root/siton-data-zhangmingData/YHJ/detectron2/croptrain/datasets/VisDrone/VisDrone2019-DET_val_coco.json [06/23 17:17:17 d2.data.common]: Serializing the dataset using: <class 'detectron2.data.common._TorchSerializedList'> [06/23 17:17:17 d2.data.common]: Serializing 548 elements to byte tensors and concatenating them all ... [06/23 17:17:17 d2.data.common]: Serialized dataset takes 1.45 MiB [06/23 17:17:17 d2.data.dataset_mapper]: [DatasetMapper] Augmentations used in inference: [ResizeShortestEdge(short_edge_length=(1200, 1200), max_size=1999, sample_style='choice')] INFO:croptrain.engine.inference:Start inference on 548 batches INFO:croptrain.engine.inference:Inference done 11/548. Dataloading: 0.0016 s/iter. Inference: 0.0860 s/iter. Eval: 0.0009 s/iter. Total: 0.0885 s/iter. ETA=0:00:47 INFO:croptrain.engine.inference:Inference done 74/548. Dataloading: 0.0020 s/iter. Inference: 0.0776 s/iter. Eval: 0.0008 s/iter. Total: 0.0805 s/iter. ETA=0:00:38 INFO:croptrain.engine.inference:Inference done 132/548. Dataloading: 0.0021 s/iter. Inference: 0.0781 s/iter. Eval: 0.0032 s/iter. Total: 0.0834 s/iter. ETA=0:00:34 INFO:croptrain.engine.inference:Inference done 190/548. Dataloading: 0.0021 s/iter. Inference: 0.0784 s/iter. Eval: 0.0039 s/iter. Total: 0.0845 s/iter. ETA=0:00:30 INFO:croptrain.engine.inference:Inference done 253/548. Dataloading: 0.0021 s/iter. Inference: 0.0781 s/iter. Eval: 0.0031 s/iter. Total: 0.0834 s/iter. ETA=0:00:24 INFO:croptrain.engine.inference:Inference done 309/548. Dataloading: 0.0021 s/iter. Inference: 0.0787 s/iter. Eval: 0.0038 s/iter. Total: 0.0846 s/iter. ETA=0:00:20 INFO:croptrain.engine.inference:Inference done 370/548. Dataloading: 0.0021 s/iter. Inference: 0.0788 s/iter. Eval: 0.0033 s/iter. Total: 0.0842 s/iter. ETA=0:00:14 INFO:croptrain.engine.inference:Inference done 427/548. Dataloading: 0.0021 s/iter. Inference: 0.0788 s/iter. Eval: 0.0038 s/iter. Total: 0.0847 s/iter. ETA=0:00:10 INFO:croptrain.engine.inference:Inference done 490/548. Dataloading: 0.0021 s/iter. Inference: 0.0785 s/iter. Eval: 0.0034 s/iter. Total: 0.0841 s/iter. ETA=0:00:04 INFO:croptrain.engine.inference:Total inference time: 0:00:45.689610 (0.084143 s / iter per device, on 1 devices) INFO:croptrain.engine.inference:Total inference pure compute time: 0:00:42 (0.078699 s / iter per device, on 1 devices)