Windaway / FCRN-Depth-Prediction-Tensorflow

Another FCRN for Depth Prediction implement with Tensorflow and Tensorlayer.
GNU General Public License v2.0
11 stars 4 forks source link

how to use training code? #7

Open abbaskali opened 2 years ago

abbaskali commented 2 years ago

how to use training code, I used fcrn_train.py, but no any output gives!

Windaway commented 2 years ago

You need to use gen_tfrecord.py to convert data to tfrecord, then use fcrn_train.py to train. Because you don't have valid train.tfrecords and test.tfrecords, so the codes can't run correctly.

abbaskali commented 2 years ago

@Windaway thank you I did like your instructions gen_tfrecord.py works good, but fcrn_train.py still doesn't work, what is the TensorFlow version should I configure, CUDA, and CUDNN? i think in configurations there is an error

Windaway commented 2 years ago

The codes may works on tensorflow 1.10.

abbaskali commented 2 years ago

@Windaway thank you but TensorLayer does not support the Tensorflow versions older than 2.0.0.!! how I can solve this!

Windaway commented 2 years ago

You may install tensorlayer==1.10.0

abbaskali commented 2 years ago

@Windaway it is working thank you, there are some errors and solved it, thankful to you, 谢谢您

abbaskali commented 2 years ago

@Windaway today is the third day for training is it normal or something wrong? I used 1348 RGB images and 1348 images corresponding depth for training and 100 RGB images and 100 images corresponding depth for testing using GPU with GTX 1070 with 8GB

ahuzsg commented 2 years ago

how should I prepare or get those train datasets? NYU-data is .mat! which type should be?