TridentNet

Last updated on Feb 23, 2021

TridentNet (R-50, 1x, caffe, MS train=N)

lr sched 1x
Backbone Layers 50
File Size 129.37 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time

Architecture TridentNet Block, ResNet, Soft-NMS
MS train N
lr sched 1x
Backbone Layers 50
SHOW MORE
SHOW LESS
TridentNet (R-50, 1x, caffe, MS train=Y)

lr sched 1x
Backbone Layers 50
File Size 129.37 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time

Architecture TridentNet Block, ResNet, Soft-NMS
MS train Y
lr sched 1x
Backbone Layers 50
SHOW MORE
SHOW LESS
TridentNet (R-50, 3x, caffe, MS train=Y)

lr sched 3x
Backbone Layers 50
File Size 129.37 MB
Training Data MS COCO
Training Resources 8x NVIDIA V100 GPUs
Training Time

Architecture TridentNet Block, ResNet, Soft-NMS
MS train Y
lr sched 3x
Backbone Layers 50
SHOW MORE
SHOW LESS
README.md

Scale-Aware Trident Networks for Object Detection

Introduction

[ALGORITHM]

@InProceedings{li2019scale,
  title={Scale-Aware Trident Networks for Object Detection},
  author={Li, Yanghao and Chen, Yuntao and Wang, Naiyan and Zhang, Zhaoxiang},
  journal={The International Conference on Computer Vision (ICCV)},
  year={2019}
}

Results and models

We reports the test results using only one branch for inference.

Backbone Style mstrain Lr schd Mem (GB) Inf time (fps) box AP Download
R-50 caffe N 1x 37.7 model | log
R-50 caffe Y 1x 37.6 model | log
R-50 caffe Y 3x 40.3 model | log

Note

Similar to Detectron2, we haven't implemented the Scale-aware Training Scheme in section 4.2 of the paper.

Results

Object Detection
BENCHMARK MODEL METRIC NAME METRIC VALUE GLOBAL RANK
COCO minival TridentNet (R-50, 3x, caffe, MS train=Y) box AP 40.3 # 71
COCO minival TridentNet (R-50, 1x, caffe, MS train=N) box AP 37.7 # 97
COCO minival TridentNet (R-50, 1x, caffe, MS train=Y) box AP 37.6 # 98