loganwilliams

import xml.etree.ElementTree as ET
  
def get_pano(pano_id, api_key=your_api_key_here):
    url_request = "https://maps.googleapis.com/maps/api/streetview?size=320x320&pano=" + pano_id + "&fov=120&heading=0&pitch=90&key=" + api_key
    response = requests.get(url_request)
    img = Image.open(StringIO(response.content))
    return img

# undocumented API lets us get more pano metadata
def outdoors(lat, lon):

loganwilliams

import shutil
import os.path
for i in range(len(pairs)):
    if i in train_set:
        s = 'train'
    elif i in test_set:
        s = 'test'
    elif i in val_set:
        s = 'val'
    

loganwilliams

function defineG(input_nc, output_nc, ngf)
    local netG = nil
    if     opt.which_model_netG == "encoder_decoder" then netG = defineG_encoder_decoder(input_nc, output_nc, ngf)
    elseif opt.which_model_netG == "unet" then netG = defineG_unet(input_nc, output_nc, ngf)
    elseif opt.which_model_netG == "unet_128" then netG = defineG_unet_128(input_nc, output_nc, ngf)
    elseif opt.which_model_netG == "unet_512" then netG = defineG_unet_512(input_nc, output_nc, ngf)
    else error("unsupported netG model")
    end
   
    netG:apply(weights_init)

loganwilliams

function defineG_unet_512(input_nc, output_nc, ngf)
    -- Two layer more than the default unet to handle 512x512 input
    local netG = nil
    -- input is (nc) x 512 x 512
    local e1 = - nn.SpatialConvolution(input_nc, ngf, 4, 4, 2, 2, 1, 1)
    -- input is (ngf) x 256 x 256
    local e2 = e1 - nn.LeakyReLU(0.2, true) - nn.SpatialConvolution(ngf, ngf * 2, 4, 4, 2, 2, 1, 1) - nn.SpatialBatchNormalization(ngf * 2)
    -- input is (ngf * 2) x 128 x 128
    local e3 = e2 - nn.LeakyReLU(0.2, true) - nn.SpatialConvolution(ngf * 2, ngf * 4, 4, 4, 2, 2, 1, 1) - nn.SpatialBatchNormalization(ngf * 4)
    -- input is (ngf * 4) x 64 x 64

loganwilliams

from StringIO import StringIO
import requests

def get_image(lat, lon, api_key=your_api_key_here):
    url_request = "https://maps.googleapis.com/maps/api/streetview?size=320x320&location=" + str(lat) + "," + str(lon) + "&fov=120&heading=0&pitch=90&key=" + api_key
    response = requests.get(url_request)
    img = Image.open(StringIO(response.content))
    return img

loganwilliams

-- usage: DATA_ROOT=/path/to/data/ name=expt1 which_direction=BtoA th hires_generate.lua
--
-- code derived from https://github.com/soumith/dcgan.torch
--

require 'image'
require 'nn'
require 'nngraph'

util = paths.dofile('util/util.lua')

loganwilliams

# makes a grid_size x grid_size mosaic
grid_size = 100
# lat lon bounds
bounds = [-73.9922010899,40.7610814675,-73.9758396149,40.7722462525]

composite = Image.new('RGB', (320*grid_size, 320*grid_size), (0, 0, 0))

lats = np.linspace(bounds[1], bounds[3], grid_size)
lons = np.linspace(bounds[2], bounds[0], grid_size)

loganwilliams

[
  "#origin#": [
    "#howwhat.capitalize# did you feel #beforeafter# you #took# #this# #picture#?",
    ...
  ],
  "howwhat": ["how", "what"],
  "beforeafter": ["before", "after", "just before", "just after"],
  "took": ["took", "captured", "photographed", "shot", "recorded"],
  "picture": ["picture", "photo", "image", "snapshot", "pic", "photograph", "shot"],
  ...

loganwilliams

features
	areas
		building : building
		playa : natural = mud
		national park : protection_title = "National Park"
		national preserve : protection_title = "National Preserve"
		national monument : protection_title = "National Monument"
		military : landuse = military
		wilderness : protection_title = "Wilderness Area"
		lake : natural = water

loganwilliams

let calculateOpacity=(pos) => {
  // pos 0.0 - 0.1: linear ramp opacity 0.0 -> 0.8
  // pos 0.1 - 0.9: opacity 0.8
  // pos 0.9 - 1.0: linear ramp opacity 0.8 -> 0.0
  if (pos < 0.1) {
    opacity = pos * 8;
  } else if (pos >= 0.9) {
    opacity = (1.0 - pos) * 8;
  } else {
    opacity = 0.8;