Method

Faster RCNN [Faster RCNN]


Submitted on 25 Apr. 2017 07:32 by
Prithviraj Dhar (Nanyang Technological University, Singapore)

Running time:0.11 s
Environment:GPU @ 2.5 Ghz (Python)

Method Description:
Trained Faster RCNN

(Still working on it)
Parameters:
iterations: 400000
base_lr: 1e-3
lr_policy: "step"
gamma: 0.1
stepsize: 50000
display: 20
average_loss: 100
# iter_size: 1
momentum: 0.9
weight_decay: 0.0005
Latex Bibtex:

Detailed Results

Object detection and orientation estimation results. Results for object detection are given in terms of average precision (AP) and results for joint object detection and orientation estimation are provided in terms of average orientation similarity (AOS).


Benchmark Easy Moderate Hard
Car (Detection) 62.31 % 56.58 % 45.27 %
This table as LaTeX


2D object detection results.
This figure as: png eps pdf txt gnuplot




eXTReMe Tracker