Skip to content

Object detection tools installation

Brian Wandell edited this page Jun 22, 2018 · 6 revisions

Installation

We use network models from the tensorflow site, and particularly sub-directory 'object_detection'. When we use models from that project we cite

Speed/accuracy trade-offs for modern convolutional object detectors.
Huang J, Rathod V, Sun C, Zhu M, Korattikara A, 
  Fathi A, Fischer I, Wojna Z, Song Y, Guadarrama S, Murphy K
CVPR 2017

We are installing access to these Tensorflow methods within isetcloud, though originally HB did it differently (see below).

Protoc issue

We had to have protoc version 3.3.0 installed on gray for the object_detection code to run. The issue and solution is described here.

https://github.com/tensorflow/models/issues/1834

On gray, we moved the installed protoc to protoc-2.5.0 and made protoc a link to protoc_3.3/bin/protoc.

Notes

HB uses the object methods in the TensorFlow site to evaluate different sensors. He forked the TensorFlowModels repository. The fork has two changes to run with our project.

  • The first is an added .py script to generate tfRecord files for our data sets.
  • The second was to modify a function to output object predictions in a JSON file.

The installation instructions for object_detection repository can be found here. There are a large number of pip installs to get this set up, but we are now doing this on GCP using vistalab VM. HB did these on gray.

Clone this wiki locally