Tf-Rec | Tensorflow 2 to utilize GPU Acceleration | Recommender System library

I was trying to save images of different sizes into tf-records. I found that even though the images have different sizes, I can still load them with FixedLenFeature.

By checking the docs on FixedLenFeature and VarLenFeature, I found the difference seems to be that VarLenFeauture returns a sparse tensor.

Could anyone illustrate some situations one should use FixedLenFeature or VarLenFeature?

I am getting used to the new dataset API and try to do some time series classification. I have a dataset formatted as tf-records in the shape of: (time_steps x features). Also I have a label for each time step. (time_steps x 1)

What I want to do is to reformat the dataset to have a rolling window of time steps like this: (n x windows_size x features). With n being the amounts of time_steps-window_size (if I use a stride of 1 for the rolling window)

The labels are supposed to be (window_size x 1), meaning that we take the label of the last time_step in the window.

I already know, that I can use tf.sliding_window_batch() to create the sliding window for the features. However, the labels get shaped in the same way, and I do not know how to do this correctly: (n x window_size x 1

How do I do this using the tensorflow dataset API? https://www.tensorflow.org/programmers_guide/datasets

Thanks for your help!

This is a follow up to these two SO questions

Select random value from row in a TF.record array, with limits on what the value can be?

Numpy array to TFrecord

The first one mentioned that tfrecords can handle variable length data using tf.VarLenFeature()

I am still having trouble figuring how how to convert my number array to tfrecord file though. Here's what the first 10 rows looks like

I have about 1.7 million observations. Each of them has about 4000 boolean features and 4 floating point labels/targets. The features are sparse and approximately homogeneously distributed (about 150 of the 4000 boolean values are set to True per observation).

If I store the whole (1700000, 4000) matrix as raw numpy file (npz format), it takes about 100 MB of disk space. If I load it via np.load(), it takes a few minutes and my RAM usage rises by about 7 GB, which is fine on its own.

The problem is, that I have to turn over my boolean values in a feed_dict to a tf.placeholder in order for the tf.data.Dataset to be able to use it. This process takes another 7 GB of RAM. My plan is to collect even more data in the future (might become more than 10 million observations at some point).

Question: So how can I feed the data to my DNN (Feed-Forward, dense, not convolutional and not recurrent) without creating a bottle-neck and in a way that is native to TensorFlow? I would have thought that this is a pretty standard setting and many people should have that problem – why not? What do I do wrong/different than people without the problem?

I heard the tfrecord format is well integrated with TensorFlow and is able to load lazy but I think it is a bad idea to use that format for my feature structure as it creates one Message per observation and saves the features as map with the keys of all features as string per observation.

I am trying to create data batches for training 2 class semantic segmentation network. The target segmented image has 2 layers, first layer with 1 for all pixels of class-1 and 0 otherwise. The second layer has pixels inverted.

In the dataset I have the output images are 3 channel rgb images with [255,255,255] and [0,0,0]. The input and output images are stored in tf-record files.

When I was experimenting in numpy I created a 2 channel binary image with code below:

I am using ubuntu 16.04, with GPU Geforce 1080, 8 GB GPU memory.

I have properly created TF-record files, and I trained the model successfully. However I still have two problems.

I did the following steps and I still have two problems, just tell me please what I am missing:-

I used VOCdevkit and I properly created two files which are:- pascal_train.record and pascal_val.record

Then,

1- From this link, I used the raccoon images, I placed them into the following directory models/object_detection/VOCdevkit/VOC2012/JPEGImages (after I deleted the previous images).

Then, I used the raccoon annotation, I placed them into the following directory models/object_detection/VOCdevkit/VOC2012/Annotation (after I deleted the previous ones).

2- I modified the models/object_detection/data/pascal_label_map.pbxt and I wrote one class name which is 'raccoon'

3- I used ssd_mobilenet_v1_pets.config. I modified it, the number of class is only one and I did not train from scratch, I used ssd_mobilenet_v1_coco_11_06_2017/model.ckpt

I am trying to get deterministic behaviour from tf.train.shuffle_batch(). I could, instead, use tf.train.batch() which works fine (always the same order of elements), but I need to get examples from multiple tf-records and so I am stuck with shuffle_batch().

I am using:

As denoted here youtube-8m tf-records are saved with the format comes at the end of my question.I write a code to extract features. but there is a problem. the code can read all elements in features successfully but it is not able to read feature_lists. in fact, the example does not include features_list at all and I get an error while I try to access it. How can I read the feauures_list. I attach Data format, My code and the output :

I'm using slim to convert data into TF-Record format and looking at this example, where the MNIST data-set is being converted.

On lines 127 to 128, the image png_string is assigned a label, labels[j].