Mask_rcnn: Detection Time (Question)

[iliachigogidze]

Hey,

When I run Detection from balloon_model on average it takes 25 sec. Why is it so slow? Because of my PC?

P.S
I don't want to spam this repo with stupid questions. I will delete this one in a day or two :)

image_id = random.choice(dataset.image_ids)
image, image_meta, gt_class_id, gt_bbox, gt_mask =\
    modellib.load_image_gt(dataset, config, image_id, use_mini_mask=False)
info = dataset.image_info[image_id]
print("image ID: {}.{} ({}) {}".format(info["source"], info["id"], image_id, 
                                       dataset.image_reference(image_id)))

# Run object detection
results = model.detect([image], verbose=1)

# Display results
ax = get_ax(1)
r = results[0]
visualize.display_instances(image, r['rois'], r['masks'], r['class_ids'], 
                            dataset.class_names, r['scores'], ax=ax,
                            title="Predictions")
log("gt_class_id", gt_class_id)
log("gt_bbox", gt_bbox)
log("gt_mask", gt_mask)

[rgommers]

Probably, 25sec sounds normal for an average PC without GPU. You need a machine with a NVIDIA GPU to speed that up by a lot; with a GTX1080Ti it shound run in under a second per step.

Your question is not Mask-RCNN specific, I suggest you close it and look into running Tensorflow on a GPU.

[iliachigogidze]

Sure, I will close it.
To clarify my question... I'm not saying about training time. I have trained model on cloud and with created model I want to test model with validation images on my i3 PC and for 1 image it takes 25 sec to detect my custom object in the image. Is that normal?

Thanks

[rgommers]

Ah no, inference should be much faster. But loading an image could be slow if it's large maybe. You probably should do a bit of profiling to measure where the time is actually spent.

[badawa]

Probably, 25sec sounds normal for an average PC without GPU. You need a machine with a NVIDIA GPU to speed that up by a lot; with a GTX1080Ti it shound run in under a second per step.

Your question is not Mask-RCNN specific, I suggest you close it and look into running Tensorflow on a GPU.

the 'step' u mentioned here means per picture?