Skip to content

Latest commit

 

History

History
56 lines (46 loc) · 4.85 KB

MODEL_ZOO.md

File metadata and controls

56 lines (46 loc) · 4.85 KB

SimpleDet Model Zoo

Introduction

This file documents a large collection of baselines trained with SimpleDet.

Common Settings

  • All models were trained on train2014+valminusminival2014, and tested on minival2014.
  • We adopt the same training schedules as Detectron. 1x indicates 6 epochs and 2x indicates 12 epochs since we append flipped images into training data.
  • We report the training GPU memory as what nvidia-smi shows.

ImageNet Pretrained Models

We provide the ImageNet pretrained models used by SimpleDet. Unless otherwise noted, these models are trained on the standard ImageNet-1k dataset.

  • resnet-v1-50: converted copy of MARA's original ResNet-50 model.
  • resnet-v1-101: converted copy of MARA's original ResNet-101 model.
  • resnet-50: ResNet-v2-50 model provided by MXNet Model Gallery.
  • resnet-101: ResNet-v2-101 model provided by MXNet Model Gallery.

Box, and Mask Detection Baselines

All AP results are reported on minival2014 of the COCO dataset.

Model Backbone Head Train Schedule GPU Image/GPU FP16 Train MEM Train Speed Box AP(Mask AP) Link
Faster R50v1-C4 C5-512ROI 1X 8X 1080Ti 2 no 8.4G 20 img/s 34.2 model
Faster R50v1-C4 C5-512ROI 1X 8X TitanV 2 yes 6.1G 49 img/s 34.4 model
Faster R50v2-C4 C5-256ROI 1X 8X 1080Ti 2 no 5.1G 33 img/s 32.8 model
Cascade R50v2-C5 2MLP 1X 8X 1080Ti 2 no 5.3G 27 img/s 37.5 model
Faster R50v1-FPN 2MLP 1X 8X 1080Ti 2 no 5.2G 36 img/s 36.5 model
Mask R50v1-FPN 2MLP+4CONV 1X 8X 1080Ti 2 no 6.7G 19 img/s 37.1(33.7) model
Retina R50v1-FPN 4Conv 1X 8X 1080Ti 2 no 5.1G 44 img/s 35.6 model
Trident R50v2-C4 C5-128ROI 1X 8X 1080Ti 2 no 7.2G 19 img/s 36.4 model
Faster R101v2-C4 C5-256ROI 1X 8X 1080Ti 2 no 6.7G 25 img/s 37.6 model
Faster-SyncBN R101v2-C4 C5-256ROI 1X 8X 1080Ti 2 no 7.8G 17 img/s 38.6 model
Faster R101v1-C4 C5-512ROI 1X 8X 1080Ti 2 no 10.2G 16 img/s 38.3 model
Faster R101v1-C4 C5-512ROI 1X 8X TitanV 2 yes 7.0G 35 img/s 38.1 model
Faster R101v1-FPN 2MLP 1X 8X 1080Ti 2 no 7.5G 24 img/s 38.7 model
Cascade R101v2-C5 2MLP 1X 8X 1080Ti 2 no 7.1G 23 img/s 40.0 model
Trident R101v2-C4 C5-128ROI 1X 8X 1080Ti 1 no 6.6G 9 img/s 40.6 model
Trident-Fast R101v2-C4 C5-128ROI 1X 8X 1080Ti 1 no 6.6G 9 img/s 39.9 model
Retina R101v1-FPN 4Conv 1X 8X 1080Ti 2 no 7.1G 31 img/s 37.8 model

FP16 Speed Benchmark

Here we provide the FP16 speeed benchmark results of several models.

Model Backbone Head Train Schedule GPU Image/GPU FP16 Train MEM Train Speed
Faster R50v1-C4 C5-512ROI 1X 8X 1080Ti 2 no 8.4G 20 img/s
Faster R50v1-C4 C5-512ROI 1X 8X TitanV 2 yes 6.1G 49 img/s
Faster R50v1-C4 C5-512ROI 1X 8X TitanV 4 yes 11.2G 55 img/s
Faster R50v2-C4 C5-256ROI 1X 8X 1080Ti 2 no 5.1G 33 img/s
Faster R50v2-C4 C5-256ROI 1X 8X TitanV 2 yes 3.8G 61 img/s
Faster R50v2-C4 C5-256ROI 1X 8X TitanV 4 yes 6.6G 73 img/s
Faster R101v1-C4 C5-512ROI 1X 8X 1080Ti 2 no 10.2G 16 img/s
Faster R101v1-C4 C5-512ROI 1X 8X TitanV 2 yes 7.0G 35 img/s