Updated goals

TODO.goals

@@ -1,25 +1,20 @@
-# US/Pacific 7/7/2014
+# US/Pacific 7/23/2014
 = Deploy web app
     = Write dummy web app for uploading a request and downloading the response
-        = Finish crosscompute-scaffolds
-    Write dummy queue that takes a request and returns a response
     Write dummy processor that takes a request and returns a response
     Replace dummy proccesor with real processor
+    Feature Modi Research Group logo and link on tool webpage
 Improve development classifier
-    Investigate why the positive count is 7027 instead of 7028
     Consider optimizing get_counts by searching maximums locally
     Consider adding random caching to get_next_batch
     Consider making coarse classifier to optimize speed
-    Experiment with training on test set
     Experiment with autoencoders to pretrain neural network
-    Experiment with YUV colorspace
+    Use cuda-convnet2
 Combine Amazon Mechanical Turk results into a clean dataset [8h]
     Set up Amazon Mechanical Turk to estimate the building count in a window [10h]
 Make scripts easier to use
     Let options.cfg read environment variables
 Send access credentials to all tools, valid with no membership fee for one year
     Add payment
-    Experiment with message queues
     Mask paths in run.pkl
-    Feature Modi Research Group logo and link on tool webpage
-Evaluate classifiers on cost-effective lower resolution satellite images
+Evaluate classifiers on cost-effective lower resolution satellite images

TODO.log

@@ -1,4 +1,7 @@
-# UTC 07/07/2014
+# UTC 07/23/2014
++ Investigate why the positive count is 7027 instead of 7028 [07/23/2014]
++ Prepare small test geotiff [07/23/2014]
++ Finish crosscompute-scaffolds [07/23/2014]
 + Select production classifier [07/07/2014]
 + Experiment with boto [07/07/2014]
 + Evaluate classifiers on myanmar0 [07/04/2014]