visualization

[YashwanthReddy27]

how do i provide input to the model so that i get the image or video with bboxes with class labels(how can i visualize using a dataset of my own)
i run the validation file with the gen1 data i get only the log files but no image or video file playing showing the labels or bounding boxes

[magehrig]

At the moment, you have to write the code to do this yourself. Currently, the repo only provides the code to replicate the results in the paper without any add-ons.

But your request is reasonable, and it would be nice to visualize the output. I hope that I can provide an example soon. In the meantime, you can have a look at the code that already logs predictions for a start.

[YashwanthReddy27]

Thank you magehrig for considering the request

[YashwanthReddy27]

when we run validation.py I'm getting the log files so are those given from the code that you mentioned here or can you please explain on how we can use or pass arguments to detection.py to get the logs predictions

[magehrig]

the logging happens in the pytorch lightning callback which is not used in validation.py. The following could work for you:

• add the visualization callback to the list of callbacks in the validation.py script. Here it is done in the train.py script.
• Instead of logging to the logger here, write your own code that writes to disk.

I cannot help you more at this point because I don't have much time. I appreciate your understanding.

[YashwanthReddy27]

Thank you for taking your time and answering the question have a great day!

[YashwanthReddy27]

Hi magehrig i do not have much knowledge on pytorch lightning i'm going through the documentation but i try to run the prediction from trainer but i'm getting this error could you please help the problem is with the dataloader could you tellme how did you call val_dataloader from lightning module i tried the similaar way as you it is calling predict dataloader but yet i'm facing this error

[magehrig]

According to the error message, you are passing a dict as input to the network (the first operation is a conv which throws this error) instead of a Tensor.

[YashwanthReddy27]

Okay thank you magehrig

[YashwanthReddy27]

Can you tell how the iterations are being calculated like how many iterations are needed for one event and what one iteration is doing everytime

[YashwanthReddy27]

hi magehrig in one of the videos that you published in youtube it was detecting objects in events can you tell on how you guys achieved it on how your are detecting objects in EVENTS by playing the event file
could you please shed some light on how to achieve this output using the dataset you mentioned

[YashwanthReddy27]

hey magehrig could you please consider answering this please