https://github.com/juanabascal/places365

Using places365 pre-trained Tensorflow model converted from Caffe
https://github.com/juanabascal/places365

Last synced: about 2 months ago
JSON representation

Using places365 pre-trained Tensorflow model converted from Caffe

• Host: GitHub
• URL: https://github.com/juanabascal/places365
• Owner: juanabascal
• Created: 2018-02-21T16:08:21.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2018-03-05T10:29:52.000Z (over 6 years ago)
• Last Synced: 2024-07-10T11:01:12.944Z (3 months ago)
• Language: Python
• Homepage:
• Size: 14.6 KB
• Stars: 6
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# TF Places365

Using places365 pre-trained models in TensorFlow. Previously converted from Caffe to TensorFlow format.

## Dependencies

You must need to have the following packages installed in order to run the predictor.

* Python 2.7

* TensorFlow

* Numpy

* Pillow

* resizeimage

## Getting the model

You will need to download the places365 model in Caffe from their [GitHub repository](https://github.com/CSAILVision/places365). You will need both the deploy file and the weights file. Moreover, you can find here the labels for the different categories of places365.

After that you need to convert the Caffe model to a TensorFlow model using the [caffe-tensorflow](https://github.com/ethereon/caffe-tensorflow) converter. You can find a guide to use the conversor [here](https://docs.google.com/document/d/1UhPTyDTFJHDx94yDH8x_dCnNpM9K9irAKZSvn97ps6c/edit?usp=sharing).

## Running

For running the program you will need to declare the image path, the converted weights paths and the labels path. Note that the input's width and height are set for a VGG16 model, you may have to adjust them to your model's input.