In [1]:

    

import mltoolbox.image.classification as model
from google.datalab.ml import *

worker_dir = '/content/datalab/tmp/coast'
preprocessed_dir = worker_dir + '/coast300'
model_dir = worker_dir + '/model300'


    

In [2]:

    

train_set = BigQueryDataSet('SELECT image_url, label FROM coast.train WHERE rand() < 0.04')
model.preprocess(train_set, preprocessed_dir)


    

    




/usr/local/lib/python2.7/dist-packages/apache_beam/coders/typecoders.py:136: UserWarning: Using fallback coder for typehint: Any.
  warnings.warn('Using fallback coder for typehint: %r.' % typehint)
WARNING:root:Dataset does not exist so we will create it
WARNING:root:Couldn't find python-snappy so the implementation of _TFRecordUtil._masked_crc32c is not as fast as it could be.






    




completed

In [3]:

    

import logging
logging.getLogger().setLevel(logging.INFO)
model.train(preprocessed_dir, 30, 1000, model_dir)
logging.getLogger().setLevel(logging.WARNING)


    

    




INFO:tensorflow:global_step/sec: 0






    




INFO:tensorflow:global_step/sec: 0
INFO:root:Train [master/0], step 1 (0.077 sec) 13.0 global steps/s, 13.0 local steps/s
INFO:root:Eval, step 1:
- on train set loss: 2.510, accuracy: 0.383
-- on eval set loss: 2.549, accuracy: 0.367
INFO:root:Eval, step 1000:
- on train set loss: 0.002, accuracy: 1.000
-- on eval set loss: 2.806, accuracy: 0.367
INFO:root:Exporting prediction graph to /content/datalab/tmp/coast/model300/model






    




INFO:tensorflow:No assets to save.






    




INFO:tensorflow:No assets to save.






    




INFO:tensorflow:No assets to write.






    




INFO:tensorflow:No assets to write.






    




INFO:tensorflow:SavedModel written to: /content/datalab/tmp/coast/model300/model/saved_model.pb






    




INFO:tensorflow:SavedModel written to: /content/datalab/tmp/coast/model300/model/saved_model.pb






    




completed

In [4]:

    

tb_id = TensorBoard.start(model_dir)


    

    





TensorBoard was started successfully with pid 29461. Click here to access it.

In [5]:

    

summary = Summary(model_dir)
summary.list_events()


    

    
Out[5]:





{u'accuracy': {'/content/datalab/tmp/coast/model300/eval_set',
  '/content/datalab/tmp/coast/model300/train_set'},
 u'batch/fraction_of_450_full': {'/content/datalab/tmp/coast/model300/eval_set',
  '/content/datalab/tmp/coast/model300/train_set'},
 u'global_step/sec': {'/content/datalab/tmp/coast/model300/train'},
 u'input_producer/fraction_of_32_full': {'/content/datalab/tmp/coast/model300/eval_set',
  '/content/datalab/tmp/coast/model300/train_set'},
 u'loss': {'/content/datalab/tmp/coast/model300/eval_set',
  '/content/datalab/tmp/coast/model300/train_set'}}

In [6]:

    

summary.plot('accuracy')
summary.plot('loss')


    

In [7]:

    

# gs://tamucc_coastline/esi_images/IMG_2849_SecDE_Spr12.jpg,3B
# gs://tamucc_coastline/esi_images/IMG_0047_SecBC_Spr12.jpg,10A
# gs://tamucc_coastline/esi_images/IMG_0617_SecBC_Spr12.jpg,7
# gs://tamucc_coastline/esi_images/IMG_2034_SecEGH_Sum12_Pt2.jpg,10A
images = [
  'gs://tamucc_coastline/esi_images/IMG_2849_SecDE_Spr12.jpg',
  'gs://tamucc_coastline/esi_images/IMG_0047_SecBC_Spr12.jpg',
  'gs://tamucc_coastline/esi_images/IMG_0617_SecBC_Spr12.jpg',
  'gs://tamucc_coastline/esi_images/IMG_2034_SecEGH_Sum12_Pt2.jpg'
]
# Set show_image to True to see the images
model.predict(model_dir, images, show_image=False)


    

    




Predicting...






    
Out[7]:






  

    

      

      
image_url
      
label
      
score
    
  
  

    

      
0
      
gs://tamucc_coastline/esi_images/IMG_2849_SecD...
      
10A
      
0.965391
    
    

      
1
      
gs://tamucc_coastline/esi_images/IMG_0047_SecB...
      
10A
      
0.964431
    
    

      
2
      
gs://tamucc_coastline/esi_images/IMG_0617_SecB...
      
7
      
0.996017
    
    

      
3
      
gs://tamucc_coastline/esi_images/IMG_2034_SecE...
      
10A
      
0.954304
    
  

In [8]:

    

eval_set = BigQueryDataSet('select * from coast.eval WHERE rand()<0.1')
model.batch_predict(eval_set, model_dir, output_bq_table='coast.eval200tinymodel')


    

    




WARNING:root:Dataset does not exist so we will create it






    




completed

In [9]:

    

ConfusionMatrix.from_bigquery('select * from coast.eval200tinymodel').plot()


    

In [10]:

    

%%bq query --name accuracy
SELECT
  target,
  SUM(CASE WHEN target=predicted THEN 1 ELSE 0 END) as correct,
  COUNT(*) as total,
  SUM(CASE WHEN target=predicted THEN 1 ELSE 0 END)/COUNT(*) as accuracy
FROM
  coast.eval200tinymodel
GROUP BY
  target


    

In [11]:

    

accuracy.execute().result()


    

    
Out[11]:





    
target
correct
total
accuracy
10B
0
3
0.0
1
0
2
0.0
8B
0
3
0.0
3B
1
20
0.05
7
1
6
0.166666666667
5
2
6
0.333333333333
6B
2
8
0.25
6A
2
10
0.2
8A
4
12
0.333333333333
9
6
20
0.3
8C
7
24
0.291666666667
2A
9
24
0.375
3A
9
14
0.642857142857
10D
10
28
0.357142857143
10A
86
116
0.741379310345
    
(rows: 15, time: 3.2s,     0B processed, job: job_No42gudIFxdN-5j6N6ShaAZGXPI)

    
  

In [12]:

    

%%bq query --name logloss

SELECT feature, AVG(-logloss) as logloss, COUNT(*) as count FROM
(
  SELECT feature, CASE WHEN correct=1 THEN LOG(prob) ELSE LOG(1-prob) END as logloss
  FROM
  (
    SELECT
    target as feature, 
    CASE WHEN target=predicted THEN 1 ELSE 0 END as correct,
    target_prob as prob
    FROM coast.eval200tinymodel
  )
)
GROUP BY feature


    

In [13]:

    

FeatureSliceView().plot(logloss)


    

In [14]:

    

import shutil
import google.datalab.bigquery as bq

TensorBoard.stop(tb_id)
bq.Table('coast.eval200tinymodel').delete()
shutil.rmtree(worker_dir)


    

In [ ]: