alexattia / simpsonrecognition Goto Github PK

consider an image that has multiple Lisa in it, Will the model put detection box on each Lisa?

i am trying to train the faster rcnn one the dataset you have provided via kaggle,
while training i am geting an error:

3/100 [..............................] - ETA: 8:11 - rpn_cls: 0.0000e+00 - rpn_regr: 0.0000e+00 - detector_cls: 5.0297 - detector_regr: 0.9954 'NoneType' object has no attribute 'shape' Exception: 'NoneType' object is not subscriptable

though the training continues i am intersted to know what is this "NoneType object"
any insight about the issue ?

Thanks.

it never changes...what have i missed?

X_train, X_test, y_train, y_test = train.get_dataset(load=True)

Hi i want to add some background images to train on my own datasets, but i saw you already define the classes_count['bg'] = 0 and no related code to add background images , could you tell me which file i can modify or which variable can be added some bachground images, thanks :)

So in the "Data Processing and Learning.ipynb"

Same pictures training/testing

model, history = train.training(model, X_train, X_test, y_train, y_test, data_augmentation=True)#, callback=True, six_conv=True)

## Training in the notebook
X_train, X_test, y_train, y_test = train.get_dataset()
model, opt = train.create_model_four_conv(X_train.shape[1:])
model.compile(loss='categorical_crossentropy',
               optimizer=opt,
               metrics=['accuracy'])
model, history = train.training(model, X_train, X_test, y_train, y_test, data_augmentation=True, callback=True, six_conv=True)

​the train.training code from line 183 on "train.py"

def training(model, X_train, X_test, y_train, y_test, data_augmentation=True):

these don't match,...

with open('./config.pickle', 'rb') as f_in:
C = pickle.load(f_in)

Is it possible to convert the model into coreml afterwards?

Exception: 'NoneType' object has no attribute 'getitem'
or
Exception: 'NoneType' object is no subscriptable
or
Exception: a must be non-empty

what is the reason of these exceptions?

Do you get this exception during training? The loss decreases and training continues, but I'd like to understand where this exception comes from and if it needs to be remedied. I realize that @yhenon has deprecated the repo, but I'm hoping to get help from other folks who've used this. Thanks in advance.

hi, when I used the code and dataset, I encountered this problem: "Exception: Error when checking target: expected rpn_out_class to have shape (None, None, None, 12) but got array with shape (1, 19, 25, 24)".

What should I do?

This project uses data with only one label...
My data format is
000001.jpg
car 13 311 84 362
car 362 330 500 389
...
car 84 323 121 350
7