The module tries to Predict Missing Objects in Images.
- Generate object sequences from object data obtained from visual genome database
cd ./src/data/
python object_sequence_generator.py allObjects.txt- Move the generated list and divide data for k-fold cross validation
cp vg.lst ../word2vec/imgs_as_objs/
cd ../word2vec/
python k_fold_generator.py imgs_as_objs/vg.lst- Run word2vec and compute the accuracy
python runner.py imgs_as_objs/Sample output for a run looks like this (prediction set size = 5%):
=== Set : 0 ===
Training : COMPLETE!
Testing : RUNNING . . .
Testing : COMPLETE!
Unknown actions: 9917; Correct predictions: 475
Set Accuracy: 4.78975496622
=== Set : 1 ===
Training : COMPLETE!
Testing : RUNNING . . .
Testing : COMPLETE!
Unknown actions: 10369; Correct predictions: 572
Set Accuracy: 5.51644324429
=== Set : 2 ===
Training : COMPLETE!
Testing : RUNNING . . .
Testing : COMPLETE!
Unknown actions: 9615; Correct predictions: 434
Set Accuracy: 4.51378055122
=== Set : 3 ===
Training : COMPLETE!
Testing : RUNNING . . .
Testing : COMPLETE!
Unknown actions: 9924; Correct predictions: 485
Set Accuracy: 4.88714228134
=== Set : 4 ===
Training : COMPLETE!
Testing : RUNNING . . .
Testing : COMPLETE!
Unknown actions: 10563; Correct predictions: 520
Set Accuracy: 4.92284388905
=== Set : 5 ===
Training : COMPLETE!
Testing : RUNNING . . .
Testing : COMPLETE!
Unknown actions: 10654; Correct predictions: 481
Set Accuracy: 4.5147362493
=== Set : 6 ===
Training : COMPLETE!
Testing : RUNNING . . .
Testing : COMPLETE!
Unknown actions: 10654; Correct predictions: 576
Set Accuracy: 5.4064201239
=== Set : 7 ===
Training : COMPLETE!
Testing : RUNNING . . .
Testing : COMPLETE!
Unknown actions: 10286; Correct predictions: 521
Set Accuracy: 5.06513707953
=== Set : 8 ===
Training : COMPLETE!
Testing : RUNNING . . .
Testing : COMPLETE!
Unknown actions: 10386; Correct predictions: 421
Set Accuracy: 4.05353360293
=== Set : 9 ===
Training : COMPLETE!
Testing : RUNNING . . .
Testing : COMPLETE!
Unknown actions: 10545; Correct predictions: 497
Set Accuracy: 4.71313418682
==== FINAL STATISTICS ====
Total missing objects: 102913; Total correct predictions: 4982
ACCURACY: 4.84098218884