def get_model_api():
config = Attention_LapConfig(load=True)
model = ImageAttentionModel(config)
model.loadTrainedModel(config.restoreModel, config.restoreModelPath)
processor = OnlineProcessor(config.trainImgFile, config)
outputGenerator = OutputGenerator('/media/jwong/Transcend/VQADataset/TrainSet/trainImgPaths.txt')
def model_api(input_qn, img_path, usePreprocessedImg=False, img_id = 214587):
if usePreprocessedImg:
#/media/jwong/Transcend/VQADataset/TrainSet/Images_train2014/COCO_train2014_000000214587.jpg
#call model predict function
alpha, pred = model.solve(input_qn, str(img_id), processor)
alp_imgName = outputGenerator.getSingleOutput(alpha, img_id, input_qn, pred)
#/media/jwong/Transcend/VQADataset/TrainSet/Images_train2014/COCO_train2014_000000299333.jpg
output_data = {"input": input_qn,
"ans": pred,
"alpha": alp_imgName}
return output_data
else:
#call model predict function
alpha, pred = model.solve(input_qn, img_path, processor)
print('Prediction made: {}'.format(pred))
alp_imgName = outputGenerator.getSingleOutput(alpha, img_path, input_qn, pred)
print('Palpha made: {}'.format(alp_imgName))
output_data = {"input": 'Question: ' + input_qn,
"ans": 'Prediction: ' + pred,
"alpha": alp_imgName}
return output_data
#return lambda func
return model_api
def runtrain(args):
#config = Attention_LapConfig(load=True, args)
config = Attention_GPUConfig(load=True, args=args)
trainReader = AttModelInputProcessor(config.trainAnnotFile,
config.rawQnTrain,
config.trainImgFile,
config,
is_training=True)
valReader = AttModelInputProcessor(config.valAnnotFile,
config.rawQnValTestFile,
config.valImgFile,
config,
is_training=False)
if args.att == 'qn':
print('Attention over question and image model')
model = QnAttentionModel(config)
elif args.att == 'im':
print('Attention over image model')
model = ImageAttentionModel(config)
model.construct()
model.train(trainReader, valReader, config.logFile)
model.destruct()
trainReader.destruct()
valReader.destruct()
return config
def loadOfficialTest(args, restoreModel=None, restoreModelPath=None):
#config = Attention_LapConfig(load=True, args)
config = Attention_GPUConfig(load=True, args=args)
testReader = TestProcessor(qnFile=config.testOfficialDevQns,
imgFile=config.testOfficialImgFeatures,
config=config)
if args.att == 'qn':
print('Attention over question and image model')
model = QnAttentionModel(config)
elif args.att == 'im':
print('Attention over image model')
model = ImageAttentionModel(config)
if restoreModel is None:
model.loadTrainedModel(config.restoreModel, config.restoreModelPath)
else:
model.loadTrainedModel(restoreModel, restoreModelPath)
if restoreModelPath is None:
testOfficialResultFile = config.testOfficialResultFile
else:
testOfficialResultFile = '{}AttSubmission.json'.format(restoreModelPath)
model.runTest(testReader, testOfficialResultFile)
testReader.destruct()
print('Official test complete')
return model
def runValTest(args):
#Val set's split -- test
print('Running Val Test')
config = Attention_LapConfig(load=True, args=args)
valTestReader = TestProcessor(qnFile=config.valTestQns,
imgFile=config.valImgFile,
config=config)
#valTestReader = TrainProcessors(config.testAnnotFile,
# config.rawQnValTestFile,
# config.valImgFile,
# config,
# is_training=False)
if args.att == 'qn':
print('Attention over question and image model')
model = QnAttentionModel(config)
elif args.att == 'im':
print('Attention over image model')
model = ImageAttentionModel(config)
model.loadTrainedModel(config.restoreQnImAttModel,
config.restoreQnImAttModelPath)
model.runTest(valTestReader, 'testResFile.json')
model.destruct()
valTestReader.destruct()
def solve():
print('Running solve')
config = Attention_LapConfig(load=True, args=args)
out = OutputGenerator(config.trainImgPaths)
#img_id = raw_input('Img_id--> ')
img_id = str(262415)
img = Image.open(out.convertIDtoPath(str(img_id)))
img.show()
qn = raw_input('Question--> ')
print(qn)
model = ImageAttentionModel(config)
model.loadTrainedModel(config.restoreModel, config.restoreModelPath)
alpha, pred = model.solve(qn, img_id)
out.displaySingleOutput(alpha, img_id, qn, pred)
''' -a otest -r ./results/Att21Mar1334/att21Mar1334.meta -p ./results/Att21Mar1334/
def runVisualise():
print('Running Visuals')
config = Attention_LapConfig(load=True, args=args)
reader = AttModelInputProcessor(config.trainAnnotFile,
config.rawQnTrain,
config.trainImgFile,
config,
is_training=False)
model = ImageAttentionModel(config)
model.loadTrainedModel(config.restoreModel, config.restoreModelPath)
alphas, img_ids, qns, preds = model.runPredict(
reader, config.csvResults, 5, mini=True)
model.destruct()
reader.destruct()
out = OutputGenerator(config.trainImgPaths)
out.displayOutput(alphas, img_ids, qns, preds)
def runVisualiseVal():
print('Running Visuals')
config = Attention_LapConfig(load=True, args=args)
reader = AttModelInputProcessor(config.testAnnotFile,
config.rawQnValTestFile,
config.valImgFile,
config,
is_training=False)
model = ImageAttentionModel(config)
model.loadTrainedModel(config.restoreModel, config.restoreModelPath)
alphas, img_ids, qns, preds, labels = model.runPredict(
reader, config.csvResults, 180, mini=True, chooseBatch=0)
model.destruct()
reader.destruct()
out = OutputGenerator(config.valImgPaths)
out.displayEachSample(alphas, img_ids, qns, preds, labels, saveData=True)
def runMetricsForInternalTestSet(args, restoreModel, restoreModelPath):
print('Running metrics for model: {}'.format(restoreModel))
#config = Attention_LapConfig(load=True, args)
config = Attention_GPUConfig(load=True, args=args)
print('Running Validation Test on Model')
valTestReader = AttModelInputProcessor(config.testAnnotFile,
config.rawQnValTestFile,
config.valImgFile,
config,
is_training=False)
#restoreModel = config.restoreModel
#restoreModelPath = config.restoreModelPath
if args.att == 'qn':
print('Attention over question and image model')
model = QnAttentionModel(config)
elif args.att == 'im':
print('Attention over image model')
model = ImageAttentionModel(config)
model.loadTrainedModel(restoreModel, restoreModelPath)
lab, pred, classToAnsMap = model.runEvaluationMetrics(valTestReader)
model.destruct()
valTestReader.destruct()
#run metrics & get stats
listOfStats = runMetrics(lab, pred, classToAnsMap, restoreModelPath)
#save to pickle
data = {}
data['labels'] = lab
data['preds'] = pred
data['classToAnsMap'] = classToAnsMap
dateID = restoreModelPath.split('/')[-2]
saveToPickle(data, 'labpreds{}.pkl'.format(dateID))
print('Metrics Completed.')
return listOfStats
def validateInternalTestSet(args, model=None, restoreModelPath=None):
from vqaTools.vqaInternal import VQA
from vqaTools.vqaEval import VQAEval
#config = Attention_LapConfig(load=True, args)
config = Attention_GPUConfig(load=True, args=args)
print('Running Validation Test on Model')
valTestReader = AttModelInputProcessor(config.testAnnotFile,
config.rawQnValTestFile,
config.valImgFile,
config,
is_training=False)
if restoreModelPath is None:
restoreModel = config.restoreModel
restoreModelPath = config.restoreModelPath
if model is None:
if args.att == 'qn':
print('Attention over question and image model')
model = QnAttentionModel(config)
elif args.att == 'im':
print('Attention over image model')
model = ImageAttentionModel(config)
model.loadTrainedModel(restoreModel, restoreModelPath)
predFile = '{}PredsAtt{}.csv'.format(restoreModelPath, args.att)
results, strictAcc = model.runPredict(valTestReader, predFile)
model.destruct()
valTestReader.destruct()
print('predictions made')
vqa = VQA(config.testAnnotFileUnresolved, config.originalValQns)
vqaRes = vqa.loadRes(results, config.originalValQns)
vqaEval = VQAEval(vqa, vqaRes, n=2)
vqaEval.evaluate()
print('Writing to file..')
writeToFile(vqaEval, restoreModelPath, vqa, vqaRes, args, strictAcc)
print('Internal test complete')