

In [1]:

    
%load_ext autoreload
%autoreload 2
import re
import pandas as pd
import aphrodite.results
import vislab
import vislab.results
import vislab.datasets



In [2]:

    
label_df = vislab.datasets.pascal.get_clf_df()

Normal VOC 20-class results



In [3]:

    
results_df, preds_panel = aphrodite.results.load_pred_results(
    'pascal_oct16', vislab.config['paths']['shared_data'] + '/results', force=False)
preds_panel









    



Loaded from cache: 38 records






    Out[3]:





<class 'pandas.core.panel.Panel'>
Dimensions: 19 (items) x 17125 (major_axis) x 4 (minor_axis)
Items axis: clf pascal_class_aeroplane to clf pascal_class_tvmonitor
Major_axis axis: 2007_000027 to 2012_004331
Minor_axis axis: decaf_fc6  vw to split

Results for a single class.



In [23]:

    
selected_score = 'decaf_fc6  vw'
pred_prefix = 'clf pascal'
label = 'class_dog'

pred_df = preds_panel['{}_{}'.format(pred_prefix, label)]
pred_df['pred'] = pred_df[selected_score]
pred_df
full_binary_metrics = vislab.results.binary_metrics(
    pred_df, '{} full'.format(label), False, with_plot=True)

Results for multi-class.



In [24]:

    
mc_pred_df = preds_panel.minor_xs('decaf_fc6  vw').join(label_df)
pred_prefix = 'clf pascal'



In [52]:

    
full_mc_metrics = vislab.results.multiclass_metrics(
    mc_pred_df, pred_prefix, balanced=False,
    with_plot=True, with_print=True)









    



------------------------------------------------------------
Classification metrics on the full dataset
top_k_accuracies: [0.627, 0.791, 0.872, 0.919, 0.948, 0.963, 0.972, 0.981, 0.986, 0.99, 0.992, 0.994, 0.995, 0.997, 0.998, 0.998, 0.999, 1.0, 1.0]
                   precision    recall  f1-score  support
class_aeroplane     0.845722  0.842179  0.843947      716
class_bicycle       0.684800  0.709784  0.697068      603
class_bird          0.824087  0.753713  0.787330      808
class_boat          0.742215  0.812500  0.775769      528
class_bottle        0.528107  0.463636  0.493776      770
class_bus           0.930131  0.488532  0.640602      436
class_car           0.770690  0.478075  0.590099      935
class_cat           0.834171  0.755232  0.792741     1099
class_chair         0.650485  0.191246  0.295588     1051
class_cow           0.591912  0.501558  0.543002      321
class_diningtable   0.154143  0.615385  0.246533      130
class_dog           0.817500  0.569191  0.671113     1149
class_horse         0.574928  0.827801  0.678571      482
class_motorbike     0.708625  0.706977  0.707800      430
class_pottedplant   0.320080  0.626459  0.423684      257
class_sheep         0.594667  0.688272  0.638054      324
class_sofa          0.280967  0.657244  0.393651      283
class_train         0.537217  0.905455  0.674340      550
class_tvmonitor     0.397338  0.749104  0.519255      279
accuracy: 0.627028966012



In [53]:

    
full_mc_metrics['binary_metrics_df']









    Out[53]:






  
    
      
      accuracy
      ap
      auc
      mcc
      pr_fig
      results_df
      roc_fig
    
  
  
    
      class_aeroplane
       0.8603713
       0.8927225
       0.9859671
         0.51083
       None
       [precision, recall]
       None
    
    
      class_bicycle
       0.6527666
       0.6507277
       0.9429867
        0.279426
       None
       [precision, recall]
       None
    
    
      class_bird
        0.687472
       0.8353669
        0.969846
       0.3457068
       None
       [precision, recall]
       None
    
    
      class_boat
       0.7936508
        0.784394
       0.9732621
       0.3799048
       None
       [precision, recall]
       None
    
    
      class_bottle
       0.3895615
       0.4901989
       0.9003062
       0.1835326
       None
       [precision, recall]
       None
    
    
      class_bus
       0.8294323
       0.8513738
       0.9843531
       0.3963753
       None
       [precision, recall]
       None
    
    
      class_car
       0.6191373
       0.6574816
        0.921111
       0.3464022
       None
       [precision, recall]
       None
    
    
      class_cat
       0.6938391
       0.8213718
         0.96655
        0.395568
       None
       [precision, recall]
       None
    
    
      class_chair
       0.5925029
       0.6581116
       0.9283187
        0.343856
       None
       [precision, recall]
       None
    
    
      class_cow
       0.7318626
       0.3643397
       0.9508406
       0.2656333
       None
       [precision, recall]
       None
    
    
      class_diningtable
        0.716438
       0.6821692
       0.9559175
       0.3415181
       None
       [precision, recall]
       None
    
    
      class_dog
       0.6804771
       0.7822894
       0.9506014
       0.4072089
       None
       [precision, recall]
       None
    
    
      class_horse
       0.6812842
       0.6853829
       0.9634471
       0.2867896
       None
       [precision, recall]
       None
    
    
      class_motorbike
       0.8576809
       0.7757554
        0.973002
       0.4565035
       None
       [precision, recall]
       None
    
    
      class_pottedplant
       0.4234598
       0.5003621
       0.9111148
       0.1767972
       None
       [precision, recall]
       None
    
    
      class_sheep
         0.77437
       0.5758884
       0.9637824
       0.2984801
       None
       [precision, recall]
       None
    
    
      class_sofa
       0.5528652
       0.4769162
        0.920226
       0.2547741
       None
       [precision, recall]
       None
    
    
      class_train
       0.6713299
       0.7605917
       0.9719644
       0.2964198
       None
       [precision, recall]
       None
    
    
      class_tvmonitor
       0.6513317
       0.6938669
       0.9572594
       0.2933467
       None
       [precision, recall]
       None
    
    
      _mean
       0.6768333
       0.6810164
       0.9521503
       0.3294249
        NaN
                       NaN
        NaN

Results for the 4 aggregate metaclasses



In [60]:

    
results_df, preds_panel = aphrodite.results.load_pred_results(
    'pascal_mc_oct16', vislab.config['paths']['shared_data'] + '/results', force=True)
preds_panel









    



Results in collection pascal_mc_oct16: 8






    Out[60]:





<class 'pandas.core.panel.Panel'>
Dimensions: 4 (items) x 17125 (major_axis) x 4 (minor_axis)
Items axis: clf pascal_metaclass_animal to clf pascal_metaclass_vehicle
Major_axis axis: 2007_000027 to 2012_004331
Minor_axis axis: decaf_fc6  vw to split



In [61]:

    
mc_pred_df = preds_panel.minor_xs('decaf_fc6  vw').join(label_df)
pred_prefix = 'clf pascal'



In [63]:

    
full_mc_metrics = vislab.results.multiclass_metrics(
    mc_pred_df, pred_prefix, balanced=False,
    with_plot=True, with_print=True)









    



------------------------------------------------------------
Classification metrics on the full dataset
top_k_accuracies: [0.634, 0.925, 0.989, 1.0]
                   precision    recall  f1-score  support
metaclass_animal    0.963374  0.529758  0.683603     4419
metaclass_indoor    0.777259  0.179174  0.291217     2785
metaclass_person    0.556489  0.969392  0.707075     7449
metaclass_vehicle   0.740019  0.322411  0.449141     2472
accuracy: 0.63404379562



In [ ]:

    
full_mc_metrics['binary_metrics_df']

	accuracy	ap	auc	mcc	pr_fig	results_df	roc_fig
class_aeroplane	0.8603713	0.8927225	0.9859671	0.51083	None	[precision, recall]	None
class_bicycle	0.6527666	0.6507277	0.9429867	0.279426	None	[precision, recall]	None
class_bird	0.687472	0.8353669	0.969846	0.3457068	None	[precision, recall]	None
class_boat	0.7936508	0.784394	0.9732621	0.3799048	None	[precision, recall]	None
class_bottle	0.3895615	0.4901989	0.9003062	0.1835326	None	[precision, recall]	None
class_bus	0.8294323	0.8513738	0.9843531	0.3963753	None	[precision, recall]	None
class_car	0.6191373	0.6574816	0.921111	0.3464022	None	[precision, recall]	None
class_cat	0.6938391	0.8213718	0.96655	0.395568	None	[precision, recall]	None
class_chair	0.5925029	0.6581116	0.9283187	0.343856	None	[precision, recall]	None
class_cow	0.7318626	0.3643397	0.9508406	0.2656333	None	[precision, recall]	None
class_diningtable	0.716438	0.6821692	0.9559175	0.3415181	None	[precision, recall]	None
class_dog	0.6804771	0.7822894	0.9506014	0.4072089	None	[precision, recall]	None
class_horse	0.6812842	0.6853829	0.9634471	0.2867896	None	[precision, recall]	None
class_motorbike	0.8576809	0.7757554	0.973002	0.4565035	None	[precision, recall]	None
class_pottedplant	0.4234598	0.5003621	0.9111148	0.1767972	None	[precision, recall]	None
class_sheep	0.77437	0.5758884	0.9637824	0.2984801	None	[precision, recall]	None
class_sofa	0.5528652	0.4769162	0.920226	0.2547741	None	[precision, recall]	None
class_train	0.6713299	0.7605917	0.9719644	0.2964198	None	[precision, recall]	None
class_tvmonitor	0.6513317	0.6938669	0.9572594	0.2933467	None	[precision, recall]	None
_mean	0.6768333	0.6810164	0.9521503	0.3294249	NaN	NaN	NaN