ObjectDetect in fastai_v1

Deep Learning
1

is there any documentation on using ObjectDetect in fastai_v1?

There is a ObjectDetect dataset but how should the data be structured to use this?
https://docs.fast.ai/vision.data.html#ObjectDetectDataset

If I have a pandas frame with:

image / obj_class / top_left_x / top_left_y / width / height

how do I turn this into an ObjectDetect dataset to run the training on?

Thanks

1 Like
2

hi did you find a way to do?

3

I didn’t end up going down this route. Guess it’s been long enough now, the API docs and stuff could have change but I don’t have time to follow up right now

4

I recently adapted Jeremy’s object detection code to work on a different data set. This is described in my blog and there’s a link to my Jupyter notebook at the bottom.

1 Like