https://github.com/rpidanny/inception-embeddings-visualization

inception machine-learning python tensorboard tensorboard-visualizations tensorflow

Last synced: about 1 month ago
JSON representation

Visualize Image Embeddings on Tensorboard

Host: GitHub
URL: https://github.com/rpidanny/inception-embeddings-visualization
Owner: rpidanny
Created: 2018-10-05T15:27:49.000Z (over 6 years ago)
Default Branch: master
Last Pushed: 2018-10-18T10:03:30.000Z (over 6 years ago)
Last Synced: 2025-02-14T13:49:01.748Z (3 months ago)
Topics: inception, machine-learning, python, tensorboard, tensorboard-visualizations, tensorflow
Language: Python
Size: 438 KB
Stars: 1
Watchers: 2
Forks: 0
Open Issues: 0
Metadata Files:
- Readme: README.md

README

# Inception Embeddings Visualization

![alt text](https://github.com/rpidanny/assets/raw/master/inception-embeddings-visualization/screen.gif)

A script to generate tensorboard embeddings visualization of a group of images.

## Requirements

* python
* tensorflow
* tensorboard

## Installation

### Local

`pip install -r requierments.txt`

### Docker

Replace `tf1.8-gpu` with `tf1.8` and `nvidia-docker` with `docker` for CPU build.

#### Pull image

`docker pull rpidanny/tf1.8-gpu`

#### Run Container

Change directory to your project directory, than run in the interactive mode:

`nvidia-docker run -it -p 9001:9001 -v=$(pwd):$(pwd) -v=local_path_to_image_dir:/data --workdir=$(pwd) --rm rpidanny/tf1.8-gpu`

## Usage

### Generate Embeddings

Use `python3` while using docker.

`python main.py --image_dir=path_to_image_dir`

### Load Embeddings on Tensorboard

`tensorboard --logdir=/tmp/tensorboard_logs --port=9001`

Go to [http://localhost:9001](http://localhost:9001)