question: is there an easy way to get only class labels out for each image?

[krzysztoffiok]

Hi, very nice work!

Can i ask if there is an easy way to get only class labels out for each image?

For my project i don't need visualizations, instead i'd love a dataframe with class labels.

Thank you

[jagin]

Hi Krzysiek, I suppose that all you need is a detection mode (--config-file configs/COCO-Detection/faster_rcnn_R_50_FPN_3x.yaml) to detect classes on the frames. Then to extract class ids you can use:

predictions = data["predictions"]
instances = instances.to(self.cpu_device)
print(predictions.pred_classes)

Looking at the detectron2/utils/video_visualizer.py is the best way to know how to extract needed information from the model predictions.

[BaqirHassan]

Labels Can be shown

1. By configuring the metadata of the model by the following line of code MetadataCatalog.get(cfg.DATASETS.TRAIN[0]).thing_classes = ['class_1', 'class_2', class_3', 'and so on']

2. By adding the following argument Visualizer object "MetadataCatalog.get(cfg.DATASETS.TRAIN[0])"

The final line of code will be

        MetadataCatalog.get(cfg.DATASETS.TRAIN[0]).thing_classes = ['class_1', 'class_2', class_3', 'and so on']
        v = Visualizer(im[:, :, ::-1], MetadataCatalog.get(cfg.DATASETS.TRAIN[0]), scale=1)