Rendering 3D bounding boxes onto test images

[dpwolfe]

Hello,

Thanks for your help on the previous issue. I'm running the tools/test.py script with the intent to generate 3D bounding boxes on the test images as a result image set. I'm using the KITTI dataset. Is there a flag I can set for test.py, or use another software package, or follow some other additional steps to be able to render those 3D bounding boxes?

Also, my next step is to try using a custom generated imageset. I'll be using the pre-trained model even though it's from the perspective of a street light. They would be frames from a video capture. I was going to try to follow the structure of the KITTI dataset and see if I can repurpose the script that generated the datainfos. Do you have any suggestions for the best way to go about using a custom set of images like this?

Ultimately, it's my goal to see 3D bounding box rendering working on those images.

Thanks!

[codyreading]

Hi and thanks for the interest!

1. As of right now, I would recommend to process the detections with CaDDN to generate the result.pkl file, which contains the outputs. You can modify the visualization script from OpenPCDet in order to load in the predictions in the result.pkl file rather than from a live demo. Note that this will visualize your boxes over the point cloud, but I assume that will be okay for your purposes.

2. My recommendation is to do what you said and reformat your custom dataset in the format of KITTI. If you are not retraining on this dataset, especially due to the change of perspective, I would not expect CaDDN to work very well. You would likely have to retrain CaDDN on the custom dataset to achieve reasonable means of performance

[codyreading]

Closing due to inactivity

[dpwolfe]

Thank you @codyreading! You're right to go ahead and close this. Your advice was very helpful. As you suggested, I'm going to use that visualization code in the demo.py as a good starting point. Thanks!

[octavianplesea]

@dpwolfe can you please share the modified visualization script from OpenPCDet? Thanks!

[octavianplesea]

Hello @codyreading ,
I'm trying to do the same thing as @dpwolfe but with no success. I successfully obtain result.pkl by running test.py from CaDDN. Then, I executed their demo.py like this:

!python ../../OpenPCDet-master/tools/demo.py --cfg_file ./cfgs/kitti_models/CaDDN.yaml --ckpt ./caddn.pth --data_path ../output/kitti_models/CaDDN/default/eval/epoch_no_number/val/default/result.pkl

I've modified demo.py by modifying this method. Now it looks like this:

` def getitem(self, index):
with open(self.sample_file_list[index], 'rb') as f:
load = pickle.load(f)
points = np.asarray(load)
#else:
# raise NotImplementedError

    input_dict = {
        'points': points,
        'frame_id': index,
    }

    data_dict = self.prepare_data(data_dict=input_dict)
    return data_dict

`

Looking in the log message everything seems to work, but unfortunately I don't get any image result with the applied bounding boxes. Can you please tell me if what I'm doing wrong? I'm using Google Colab for this