deeplearning

deeplearning / Places365Classifier / 0.1.9

README.md

1. Introduction

This is an image classifier specifically trained for classifying various places. The list of places can be found here.

Input:

  • (Required) Image Data API Url, Web (http/https) Url, binary image or a base64 encoded image.
  • (Optional)?Number of results. (Default=5)

Output:

  • Top N classifications.

Note: The first call to this algorithm will take a bit longer than sequential calls to due algorithm initialization. All following calls will be significantly faster.

2. Examples

Example 1.

  • Parameter 1: Data API Url
{
    "image": "data://deeplearning/example_data/burj_khalifa.jpg"
}

Output

{
  "predictions": [
    {"class": "tower", "prob": 0.6098036170005801},
    {"class": "skyscraper", "prob": 0.14249812066555023},
    {"class": "office_building", "prob": 0.04916094988584519},
    {"class": "downtown", "prob": 0.02924365177750587},
    {"class": "church/outdoor", "prob": 0.028504755347967155}
  ]
}

Example 2.

  • Parameter 1: HTTP Url
{
    "image": "https://s3.amazonaws.com/algorithmia-assets/algo_desc_images/deeplearning_Places365Classifier/airfield.jpg"
}

Output:

{
  "predictions": [
    {"class": "airfield", "prob": 0.5347753167152406},
    {"class": "runway", "prob": 0.2845350205898285},
    {"class": "heliport", "prob": 0.09564844518899919},
    {"class": "landing_deck", "prob": 0.034962698817253106},
    {"class": "raceway", "prob": 0.02728229202330112}
  ]
}

Example 3.

  • Parameter 1: Base64 image
{
    "image": "data:image/png;base64....",
}

Output;

{
  "results":[
    {"class": "archive", "prob": 0.16057744622230533},
    {"class": "beauty_salon", "prob": 0.11514712870121004},
    {"class": "dressing_room", "prob": 0.06852224469184875},
    {"class": "biology_laboratory", "prob": 0.044164657592773444},
    {"class": "coffee_shop", "prob": 0.037719260901212685}
  ]
}

Example 4.

  • Parameter 1: Data API Url
  • Parameter 2: Number of results.
{
  "image": "data://deeplearning/example_data/coast.jpg",
  "numResults": 15
}

Output:

{
  "predictions": [
    {"class": "islet", "prob": 0.5723654627799988},
    {"class": "cliff", "prob": 0.13769967854022983},
    {"class": "coast", "prob": 0.1219320520758629},
    {"class": "ocean", "prob": 0.05200425907969475},
    {"class": "lagoon", "prob": 0.034720458090305335},
    {"class": "harbor", "prob": 0.01532892324030399},
    {"class": "lake/natural", "prob": 0.01080096885561943},
    {"class": "village", "prob": 0.008164397440850735},
    {"class": "valley", "prob": 0.005376808345317839},
    {"class": "mountain", "prob": 0.0051351888105273255},
    {"class": "river", "prob": 0.004809863399714232},
    {"class": "butte", "prob": 0.00298486975952983},
    {"class": "sky", "prob": 0.002884584246203304},
    {"class": "beach", "prob": 0.0028554031159728765},
    {"class": "rainforest", "prob": 0.002376777119934559}
  ]
}

3. Credits

For more information please refer to: Places: An Image Database for Deep Scene Understanding B. Zhou, A. Khosla, A. Lapedriza, A. Torralba and A. Oliva Arxiv, 2016

Trained model was retrieved from: Places2.

Demo image were retrieved from Wikipedia:

https://en.wikipedia.org/wiki/Skyscraper#/media/File:Burj_Khalifa.jpg

https://en.wikipedia.org/wiki/Airport#/media/File:HFX_Airport_4.jpg

https://en.wikipedia.org/wiki/Harbor#/media/File:Anacapri_BW_2013-05-14_13-32-36.jpg