How do I test the code on video data? about ssd_keras HOT 4 CLOSED

You don't need BatchGenerator to do that. Just pass the video frames to the model's predict() method.

from ssd_keras.

Thanks for your reply.
I have tried to test an image which is not belong to the validate dataset. Once, I am able to do that then I can pass the video frames into the model.predict(). I have tried as follows,
image_load=Image.open('./examples/bike.jpg')
image_load=np.asarray(image_load, dtype="float32")
test = BatchGenerator(filenames=image_load)
test_class = test.generate(batch_size=1,
train=False,
equalize=False,
brightness=False,
flip=False,
translate=False,
scale=False,
random_crop=False,
crop=False,
resize=False,
gray=False,
limit_boxes=True,
include_thresh=0.4,
diagnostics=False)

X, filenames = next(test_class)
i = 0
#Make a prediction
y_pred = model.predict(X)

The error is "ValueError: Found input variables with inconsistent numbers of samples: [1, 0]".

from ssd_keras.

I've already answered an almost identical question in #28.

Note that

1. The filenames argument in the BatchGenerator constructor expects a list that contains the paths to the images, not the images themselves. So it would have to be image_load=['./examples/bike.jpg'].
2. The first point also illustrates why it is unnecessary to use BatchGenerator here. You've already loaded the image with Image.open(), so why would you need to pass it to the batch generator? All you need to do is pass the loaded image to model.predict(). Be aware that model.predict() wants 4-dimensional input though, even if it is just one image, so if img is your image, the call has to be model.predict(np.array([img])), not model.predict(img).

from ssd_keras.

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

from ssd_keras.