Train on a custom dataset and predict bounding box on a new image.

[Rahul-Venugopal]

Hi ,

Thanks for sharing the code. Actually I am trying to use this to generate bounding boxes for my project . Is it possible to train the model on a custom data-set ? If possible , can you please tell the modifications I have to make. I am trying to see the result on a random image . Suppose if I train on custom data-set with labels (lets say I have images of cat and dogs ) , is it possible to create bounding boxes around cat on a random image ?

(PS : I am not an expert in this field and please excuse me if my questions are too vague)

Thanks Rahul

[xiaomengyc]

Hi Rahul,

It is definitely feasible to train the model on your custom dataset. You need to replace the training list in datalist and then modifying the bash scripts.

This weakly supervised localization model is doing exactly what you have described. Please refer to #2 for more details.

Best, Xiaolin

[Rahul-Venugopal]

Thanks for replying. It is actually very interesting and very helpful where I have abundant data with only image level labels . The main problem I am facing right now with implementing the model is modifications to be made in the code . I had read the paper and it would be great to read any blog or something like that , which gives more intuitive explanation about the coding part. Is there any resources currently available ?