Validation not performed

[itsron143]

Hi,
Thanks for the amazing implementation of f-rcnn. I had a doubt about the validation set. In the train_frcnn.py file I can see that the validation set has been generated but has not been used.

Here is the excerpt from train_frcnn.py:
train_imgs = [s for s in all_imgs if s['imageset'] == 'trainval']
val_imgs = [s for s in all_imgs if s['imageset'] == 'test']

print('Num train samples {}'.format(len(train_imgs)))
print('Num val samples {}'.format(len(val_imgs)))

data_gen_train = data_generators.get_anchor_gt(train_imgs, classes_count, C, nn.get_img_output_length, K.image_dim_ordering(), mode='train')
data_gen_val = data_generators.get_anchor_gt(val_imgs, classes_count, C, nn.get_img_output_length,K.image_dim_ordering(), mode='val')

data_gen_val is the last variable I see having the validation set but it has not been used during training.

Also, pascal_voc_parser.py extracts the test set as well but it has not been used for training. How can I show the validation loss and also the test loss during training?

Apologies if I got something wrong, beginner, here. Thanks for the project!

[linaemunsamy]

Hi @itsron717 - have you managed to resolve this validation problem?

[tejasri19]

@kbardool Pls Check this issue regarding validation. @itsron717 has clearly mentioned above.

[itsron143]

Hi @itsron717 - have you managed to resolve this validation problem?

Hey, I had finished working on my project so I didn't get a change to look into it.

[linaemunsamy]

Hi Guys,

I found this and managed to incorporate it into this training. It performs validation after x many epochs:
https://github.com/ant1pink/object_detection_keras/blob/master/train_frcnn.py

Hope it helps!