Skip to content

ramgendeploy/Unsplash100Labels

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

95 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Unsplash 100 multi-labels classifier

This is a classifier trained over 100 labels of unsplash photos I trained it with the fast.ai library fast.ai The labels are:

["airplane", "ambulance", "animal", "artist", "aurora", "baby", "beach", "bear", "bedroom", "bicycle", "bird", "boats", "book", "bridge", "building", "bus", "cars", "castle", "cat", "city", "clouds", "college", "column", "concert", "couple", "crops", "dance", "dawn", "deer", "desert", "dessert", "doctor", "dog", "dolphins", "field", "fire", "floor", "food", "golf", "graffiti", "grandfather", "grandmother", "grass", "hair", "hand", "horse", "hospital", "house", "human", "insect", "kid", "library", "lights", "man", "moon", "mountain", "music", "nature", "neon", "nurse", "ocean", "painting", "palm", "party", "person", "phone", "plant", "rain", "rainforest", "restaurant", "river", "robot", "rocks", "roses", "shirt", "shop", "sign", "sky", "skyscraper", "snow", "soccer", "sports", "stadium", "staircase", "stars", "storm", "street", "sun", "sunrise", "temple", "tree", "truck", "vegetable", "water", "waves", "weed", "windows", "woman", "wood"]

Heroku Guide

Rama Blog Youtube video

Partialy based on this guide

To port the fastai-Guide: Follow this guide fastai-Guide add a Procfile and put web: python app/server.py serve in it Then in the server.py file

  • import os library import os
  • Set the variable: Port = int(os.environ.get('PORT', 50000))
  • In uvicorn.run set port to Port uvicorn.run(app=app, host='0.0.0.0', port=Port, log_level="info") If heroku is not using the correct version of python, add a file runtime.txt with python-3.7.3

Data extraction and Model training

You can see the process of training and obtaining the data in this notebook, I use fast.ai library to train it and Unsplash api to get the data.

About

Classifier trained with 100 classes on unsplash images

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published