feature fusion

[250456]

How can I provide a CNN with numerical data to improve classifications using Tensorflow?
I wish to train a convolutional neural network to detect "Object X" using TensorFlow in Python.
The problem is, there is a second "Object Y" that looks identical to Object X, and the only way to differentiate between the two is to examine other sensor data.

In any case, is there a way of providing the sensor data to the CNN so that it can determine by itself if an image contains Object X or Object Y, even though they are identical (when seen in photographs)?
Or would it be easier to just use a regular CNN to get classifications, and then do an "if" function depending on the value of the sensors?
e.g. if CNN detects Object X, and sensor value is less than 30, then image contains Object Y, but if sensor value is greater than 30, then image contains Object X. Would that be easier than trying to teach a CNN how to handle both the images and sensor data?
Thanks in advance.

[AIWithShrey]

Hello, Why don’t you include the sensor data in your feature set? That way, you don’t have to write manual if conditions after you train your model.

On Sat, Jul 9, 2022 at 11:21 AM 250456 ***@***.***> wrote: How can I provide a CNN with numerical data to improve classifications using Tensorflow? I wish to train a convolutional neural network to detect "Object X" using TensorFlow in Python. The problem is, there is a second "Object Y" that looks identical to Object X, and the only way to differentiate between the two is to examine other sensor data. [image: feature fusion] <https://user-images.githubusercontent.com/101789908/178093518-5608183f-dc9d-4c9b-98c5-c880330da08a.png> [image: Screenshot 2022-07-09 111959] <https://user-images.githubusercontent.com/101789908/178093520-25c5a06d-231c-4161-980e-5d14871d68c9.png> In any case, is there a way of providing the sensor data to the CNN so that it can determine by itself if an image contains Object X or Object Y, even though they are identical (when seen in photographs)? Or would it be easier to just use a regular CNN to get classifications, and then do an "if" function depending on the value of the sensors? e.g. if CNN detects Object X, and sensor value is less than 30, then image contains Object Y, but if sensor value is greater than 30, then image contains Object X. Would that be easier than trying to teach a CNN how to handle both the images and sensor data? Thanks in advance. — Reply to this email directly, view it on GitHub <#407>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AF2IW5RGHNM6YLMG322FTXTVTEHNRANCNFSM53C27CPA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- Warmest Regards, Shreyas Mocherla Student of Grade 12, Indus International School Hyderabad Visit my page at shreyasm.com

[250456]

In which layer do i have to include CSV features and labels for better classification? Do i have to normalize it before concatenate? On Sat, Jul 9, 2022 at 3:05 PM Shreyas Mocherla ***@***.***> wrote:

…

Hello, Why don’t you include the sensor data in your feature set? That way, you don’t have to write manual if conditions after you train your model. On Sat, Jul 9, 2022 at 11:21 AM 250456 ***@***.***> wrote: > How can I provide a CNN with numerical data to improve classifications > using Tensorflow? > I wish to train a convolutional neural network to detect "Object X" using > TensorFlow in Python. > The problem is, there is a second "Object Y" that looks identical to > Object X, and the only way to differentiate between the two is to examine > other sensor data. > [image: feature fusion] > < https://user-images.githubusercontent.com/101789908/178093518-5608183f-dc9d-4c9b-98c5-c880330da08a.png > > [image: Screenshot 2022-07-09 111959] > < https://user-images.githubusercontent.com/101789908/178093520-25c5a06d-231c-4161-980e-5d14871d68c9.png > > > In any case, is there a way of providing the sensor data to the CNN so > that it can determine by itself if an image contains Object X or Object Y, > even though they are identical (when seen in photographs)? > Or would it be easier to just use a regular CNN to get classifications, > and then do an "if" function depending on the value of the sensors? > e.g. if CNN detects Object X, and sensor value is less than 30, then image > contains Object Y, but if sensor value is greater than 30, then image > contains Object X. Would that be easier than trying to teach a CNN how to > handle both the images and sensor data? > Thanks in advance. > > — > Reply to this email directly, view it on GitHub > <#407>, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/AF2IW5RGHNM6YLMG322FTXTVTEHNRANCNFSM53C27CPA > > . > You are receiving this because you are subscribed to this thread.Message > ID: ***@***.***> > -- Warmest Regards, Shreyas Mocherla Student of Grade 12, Indus International School Hyderabad Visit my page at shreyasm.com — Reply to this email directly, view it on GitHub <#407 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AYITBVDRQWWJ3QYMOGOGQ73VTFBVTANCNFSM53C27CPA> . You are receiving this because you authored the thread.Message ID: <mrdbourke/tensorflow-deep-learning/repo-discussions/407/comments/3111989@ github.com>